Backend Manager
Backend Manager follows what you have deployed, stored and monitored on 33 services, from your phone and with your own tokens. It has three tabs: Status (what needs you), Projects (every project in one list) and Services (what is connected and what you can connect). Reading is free, except live logs and the PostHog feature flags; destructive changes, rolling back, background watching and the AI explanations are Pro. Reversible actions such as redeploying or cancelling are free.
What it is and how to open it
Backend Manager opens as a full-screen window with no close button and no back arrow. The system back gesture goes up one step; from the root of a tab it returns to the welcome screen, and from the welcome screen it leaves the module. Every call goes from your phone to the provider's official API with your own token: PocketCode does not host, bill or proxy any of these services.
Ways to open it
| Where you are | What you do | Where you land |
|---|---|---|
| The navigation bar | Tools, then Backend in the Dev tools sheet. A red number on that cell means something needs attention. | The welcome screen, always. It is deliberate: the welcome screen already shows what needs you, so the module never reopens on the last tab you used. |
| The command palette | Deploy, in the Tools group. Its description still says “deploy tab”, but there are only three tabs. | Projects filtered by Hosting, with no welcome screen. |
| Home | The quick actions Deploy and Monitoring, or View all on the Backend card. | Projects filtered by Hosting or by Observability, or Status. |
| The launcher icon | Long-press PocketCode, then Status or Redeploy. | Status. Redeploy (“Redeploy last project”) never deploys by itself: it opens the screen of the project you checked most recently, or Status if there is none, where the real button and its confirmation are. |
| Quick settings and the home screen | The Backend tile and the Backend status and Backend project widgets (Pro). | Status, or the project's screen. They only open screens: they never run an action. |
| A link | pocketcode://backend/{provider}/{project}/{section} or pocketcode://backend-status. | The provider's account, the project, the section, or Status. |
| The file explorer or the editor | Deploy this project in the explorer's action menu, or Deploy Project in the editor's menu and command palette. | A sheet over the app (see Deploy from the editor). |
How the window behaves
- Each tab remembers where you were, so switching tabs or filters never loses your place. The tabs are shown only at the root of a tab.
- On a tablet or any window 600 dp wide or more, Projects and Services show a list on the left and the detail on the right; Status is limited to 720 dp wide.
- The AI assistant has its own backend tools (status, projects, logs, variables, deploy and roll back) and they follow the same Free and Pro limits as the buttons. They are described in AI Chat.
You can't, and why
- Open the module straight on a tab from the Dev tools sheet: it always starts on the welcome screen. The command palette, Home, links and shortcuts do skip it.By design
- Open a link, widget or launcher shortcut for a service other than the 13 that have a Backend Manager screen: Render, Vercel, Netlify, Railway, Fly.io, Cloudflare, DigitalOcean, Firebase, Appwrite, Convex, Supabase, Sentry and PostHog. For Neon, GitHub, Stripe and the rest the link does nothing.Not yet
- Run an action from a widget, the tile, a launcher shortcut or a link: they only open a read-only screen, where the action keeps its own confirmation and plan limits.By design
Screen map
Every screen of the module, how you get there and what you do in it.
| Screen | How you get there | What you can do |
|---|---|---|
| Welcome | Open the module from Tools; or the back gesture from the root of a tab. | Read how many services are connected and what needs you first; go to Status, Projects or Services; connect a service. |
| Header, tabs and badges | Always on top; the tabs only at the root of a tab. | Switch between Status, Projects and Services; read the red number on Status; when a provider fails, the Something went wrong card offers Retry, Copy the error and Dismiss. |
| Status | The Status tab; See what needs you; the launcher shortcut, the tile, the widgets and the Home Backend card. | Read the one-line summary, the cards of what needs attention, Platform status, Coming up, Spend and the changes you made today; ask for Summarize today (Pro). |
| Projects | The Projects tab; Open projects. | Search, filter, pull to refresh, open a project or a provider's account, check a provider you never queried. |
| Provider account | Tap a provider's name in the list; Open account in Services. | Switch between Projects and Account, open the account's sections and run its account actions. |
| Project | Tap a project (list, Status, provider account, link, widget). | See the state and the last commit, run the main action, open More actions, read the key figures and open the sections. |
| Timeline | The first activity row of a project (Deployments on hosting providers). | Filter Everything, Deploys, Errors and Config; read deploys and watch events together. |
| Section list | A section row of a project or an account; All tools; a link with a section. | Search, filter by state, Load more, copy an id, open it on the web and run the section's actions. |
| Row detail | Tap a row of a list. | Read what the provider sent, Copy all, see the steps of a CI job or the stack trace of a Sentry issue. |
| Deployment detail | Tap a deployment in Deployments. | Redeploy, Roll back to this one, Promote to production, Cancel deploy, Retry this deploy, read the Build log, open Live Logs, Explain this failure (Pro). |
| Logs | A logs section of a project; Build log in a deployment. | Search, filter by level, source, path and time, pause, copy and share. |
| Live Logs | Live Logs in a project or in a running deployment (Pro). | Follow the stream, filter, pause, copy, share and Reconnect. |
| Environment Variables | The variables section of a project. | Read the names by environment, copy a name, Add variable and Delete variable (Pro). |
| Confirmation, Pro wall and Result | Any action that changes something. | Confirm with the resource's name (and your fingerprint if the action is risky), see the wall of the Pro gate, read the Result. |
| Services | The Services tab; Connect your first service. | Search 33 services, connect one, manage a connection (see Platform Services). |
| Backend Manager notifications | Settings → Notifications → Backend Manager. | Choose per service Notify, Watch or Off, what to notify and which projects to mute (Pro). |
| Deploy this project | The file explorer's action menu or the editor. | Set a destination once, then deploy a ZIP or a commit and push. |
| Widgets, tile and shortcuts | The home screen, quick settings, a long-press on the launcher icon. | See the state at a glance and open Backend Manager; they never run an action. |
Welcome and the three tabs
The welcome screen is where the module starts when you open it from Tools. After it, everything lives in three tabs, and you go deeper by tapping: provider, project, section, detail.
The three tabs
| Tab | What it shows |
|---|---|
| Status | What needs you now: open incidents on your providers' status pages, projects that failed, the month's spend and a one-line summary. Planned maintenance is listed but never counted as an incident. |
| Projects | Every project of every connected account in one list, each with a status: Live, Deploying, Deploy failed, Degraded, Suspended or Not checked. |
| Services | Your connected services first, then the catalog of 33 to connect the rest, grouped and searchable. |
The number on Status is red and counts what needs attention; a scheduled maintenance never counts. Projects and Services carry a neutral number: real projects (not UptimeRobot monitors or CI rows) and connected services.
You can
- With no service connected, it says “Every deploy, database and error report you have, in one place.” and “33 services: hosting, databases, monitoring, payments and messaging.”, with Connect your first service, which opens Services with the search box focused.
- With services connected, it shows how many there are, the first thing that needs you and how many more want a look. Then See what needs you (to Status) or Open projects (to Projects), and Connect another service (to Services).
You can't, and why
- Choose a Deploy, BaaS or Monitoring tab: there are only three tabs. Hosting, data, observability and CI are filters inside Projects.By design
Status
Status answers one question: is anything broken? Entering it refreshes the monitors, the Expo builds and the status pages of your providers, and asks for the spend.
You can
- Read the one-line summary at the top: All good, or a mix such as “1 incident · 2 projects need you · 4 projects live”.
- Tap Summarize today for one sentence per project that needs attention (Today's summary). It is Pro and asks the AI provider you connected in PocketCode; if nothing needs attention it doesn't call the AI and says Nothing needs attention right now.
- At the top, read the attention cards (they have no heading on this tab): a project that failed or is degraded, a recent event (a new Sentry issue, a certificate about to expire, a monitor that went down, watching switched off by an invalid token) or a meter you already loaded that reached 80 % of its quota. They offer Open project, Details and Dismiss; a dismissed card comes back by itself if the thing changes (a new commit, level or impact).
- Read Platform status: open incidents on the public status pages of the providers you connected, with their impact (Minor, Major or Critical) and the project they affect. Status page opens the official site; swipe the card or tap Dismiss to hide it.
- Read Coming up: planned maintenance, grouped by provider and marked Planned. It is never counted as an incident.
- Read Spend (see the note below).
- Read Changes you made today from the phone, a local log of what you changed today (three shown, See all for the rest). Undo appears only where there is an inverse: redeploy and cancel, suspend and resume, promote and roll back.
- Tap the row of quiet projects to go to Projects, and the dismissed · Show row to bring back what you hid.
- With nothing connected it says Nothing connected yet and offers Connect a service.
You can't, and why
- See platform incidents for a service without a known public status page: only 14 are known (Vercel, Netlify, Render, Railway, Fly.io, Cloudflare, DigitalOcean, Supabase, Sentry, PostHog, GitHub, Appwrite, Convex and Neon).Third party
- See Spend for Railway, Cloudflare, Netlify or Vercel: there is no billed amount PocketCode can read for them, and Railway's usage is an estimate, which can't carry a currency.Third party
- Undo purging a cache, a DNS record, a variable or a SQL statement: only redeploy and cancel, suspend and resume, promote and roll back have an inverse.By design
- Get a card for a suspended project: you paused it on purpose, so it is not an alert.By design
What “Spend” shows
Projects
Projects is one list: your projects from every connected provider, grouped by provider with the broken ones first and then by name. Hosting, data, observability and CI are filters, not separate screens. On a tablet the list takes the left side and the detail the right (“Pick a project on the left.”).
You can
- Search by name, domain or branch (Name, domain or branch). It filters what is already loaded and asks no provider.
- Open Filter (it becomes Filter · 2 with two filters on): Category (Hosting, Data, Observability, CI) and Status (Live, Deploying, Deploy failed, Suspended, Degraded, Not checked). Data groups the backend services and the databases.
- Read each project's status stripe. CI repositories say CI passing, CI failing or Running; monitors say Up or Down.
- Tap a project to open it, or the provider's name (in capitals, with its logo) to open the account.
- Pull to refresh. It asks only the providers still visible after the filter and skips those waiting for a rate limit, which show Wait N s.
- A connected provider you never queried shows <Provider> · not checked with a Check button; one that answered with nothing shows <Provider> · no projects.
- With nothing to show you get Nothing to show yet and Connect a service; with a filter that matches nothing, No project matches that.
- Tap + (New) to open One-Click Deploy or Kubernetes. Both screens belong to the hosting stack, so opening either one also puts the list on the Hosting category.
- One-Click Deploy has three steps: Select a provider, Enter repository details (Repository URL and Branch, then Next) and Confirm deployment (Project Name, the Summary and Deploy). Only providers that can do it are listed, today Vercel: it creates the project on Vercel from the repository and ends with Deploy initiated! (“Your project is being deployed”). Without a project name and a repository (owner/name or full URL) it says “Enter a project name and a Git repository (owner/name or full URL)”.
- Kubernetes is a local generator of a Kubernetes deployment manifest, with no cluster involved: Name, Namespace, Image, Replicas, Port, CPU Limit and Memory Limit fill a YAML Preview that you can also edit by hand. Regenerate rebuilds the YAML from the fields, and the copy and share icons (Copy value, Share the manifest) copy or send it.
You can't, and why
- Filter by a single provider with a chip: the filter is by category and status.By design
- Search the providers: the search box only filters what is already loaded.By design
- Open an account screen for a CI repository or a monitor: tapping one opens its web address.By design
- Open a provider PocketCode has no screen for: you get “PocketCode has no screen for this provider yet.”Not yet
Provider account
Tap a provider's name in the list, or Open account in the Services tab, to see the account itself instead of one project.
You can
- Switch between Projects (with a count) and Account.
- In Account, open the account sections, grouped as Observe, Configure, Data, Run, Usage and More (DNS zones in Cloudflare, domains and uptime checks in DigitalOcean…), and run account actions such as Switch team on Vercel.
- Read up to three account figures.
- With no token you see Connect <Provider> and a Connect button that opens the connection sheet over the module; with no projects you see No projects found and See it in <Provider>.
You can't, and why
—
A project and its timeline
A project puts state before buttons. At the top: its name, its status and how long ago it was checked, then the last commit (“branch · sha · author”). Below: one main action and one secondary action, the key figures (up to three), Needs a look if something is pending for this project, and its sections, grouped as Observe, Configure, Data, Run, Usage and More.
You can
- Run the main action, chosen by the project's state (table below), and one secondary action. Everything else is behind More actions, under View, Change and Affects production; an action your plan doesn't open shows Pro and a lock.
- The header can offer Redeploy (Re-run on a CI repository), Roll back, Suspend or Resume (Render), Cancel deploy (only while a deployment is in flight), Live Logs and Open in browser, depending on what the provider supports. A Cloudflare project that is not a Pages project only offers to open it.
- The header Roll back goes back to the last deployment that worked: it skips the newest deployment (the one that just broke) and takes the first earlier one that was ready. It works on Vercel, Netlify, Render, Cloudflare Pages, Railway (with the service chosen for the project) and DigitalOcean; with none it says “No earlier successful deploy to go back to”. On DigitalOcean it asks DigitalOcean to roll the app back to its previous deployment, without the three phases below, which belong to Roll back to this one.
- On Netlify, Redeploy is disabled right after you use it (Wait 20 s), because Netlify allows three deployments a minute.
- Read the key figures. On Cloudflare, UptimeRobot and Railway they show Load until you tap them, to spare your request budget.
- Open All tools when a project has more than eight sections: each group then shows three rows and the sheet has a search box.
- A section row summarises its state (Asking the provider…, a count, Open it to read the logs). The first activity row opens the Timeline.
- Hide a Needs a look row of this project with Dismiss.
- Opening a project asks only for Deployments and up to three figures (and, on Cloudflare zones, the security settings); every other section is requested when you open it.
You can't, and why
- Run an irreversible action: deleting resources,
DROP, restoring backups or changes that alter your bill are never offered, here or anywhere in the module.By design - Change production without Pro: every destructive action needs the Pro gate Change your backend from the phone (see Free and Pro).Pro
- Roll back with one tap from a failed project: Roll back sits in More actions, under Change. The main action of a failed project is Why it failed (Pro) or Open build log.By design
- Load every section when you open a project: the request limit you share with the provider's own dashboard would be spent before you read anything.Third party
The main action of each state
| State | Main action | Secondary action (the first one the provider offers) and the rest |
|---|---|---|
| Deploy failed | Why it failed (Pro: it opens Deployments, where Explain this failure gives the cause). On Free it is Open build log. | Redeploy. In More actions: Open build log, Roll back, Live Logs. |
| Deploying | Live Logs (Pro: on Free, tapping it opens the Pro wall). | Cancel deploy. In More actions: Open build log and opening the site. |
| Live | Open in browser. | Redeploy. In More actions: Live Logs, Open build log. |
| Suspended | Resume. | Open build log. In More actions: Live Logs, Open in <Provider>. |
| Not checked | Check now. | Opening the site. In More actions: Live Logs, Open build log. |
| Degraded | Check now. | Live Logs. In More actions: opening the site, Roll back. |
How each kind of action asks first
| Kind of action | What happens |
|---|---|
| Read: lists, logs, metrics | Nothing to confirm. |
| Reversible: Redeploy, Re-run, Cancel deploy, Resume, Roll back, Promote to production | One confirmation that names the resource. No fingerprint. |
| Destructive: Suspend, adding or deleting variables, DNS records, domains… | The same confirmation plus your fingerprint or screen lock, if the phone can ask for it; if it can't, the confirmation alone. It needs Pro. |
Irreversible: deleting resources, DROP, restoring backups | Never offered. |
Timeline
The first activity row of a project opens the Timeline: one list that merges the project's deployments and runs with the events of background watching, newest first.
You can
- Filter with the chips Everything, Deploys, Errors and Config; only the chips that have rows appear.
- Read the list grouped by day; rows with no date go last. An empty timeline says Nothing here yet.
- Tap an event to open its link, or a deployment to open the Deployments list.
- Use the bottom bar: Tools (the project's sections) and the project's main action.
You can't, and why
- See history older than what the providers return and the phone has recorded.Third party
- See watch events if background watching has never run: they are recorded by the Pro watch.Pro
Deployments
Tap a deployment in the Deployments list of a project (Vercel, Netlify, Render, Railway, Cloudflare Pages and DigitalOcean) to see its detail: commit message, state, commit, author, branch, date, id and URL, only the fields the provider sent.
You can
- Redeploy where the provider offers it (Netlify, Render, Railway and DigitalOcean; Vercel has its own Redeploy action, which asks which deployment). On Free it works while the project has no other deployment in flight.
- Roll back to this one on a ready deployment, and Promote to production (Vercel only). They are Pro, and reversible: one confirmation, no fingerprint. On DigitalOcean Roll back to this one has three phases (below); the header Roll back doesn't use them.
- Cancel deploy while it is in flight (Vercel, Netlify, Render and Railway).
- Retry this deploy on a failed Cloudflare Pages deployment: it builds the same commit again and nothing live changes until it succeeds.
- Read the Build log (up to 20 pages) on Vercel, Render, Cloudflare and Railway.
- Open Live Logs while it is running (Pro).
- Explain this failure on a failed one (Pro). It sends an extract of the build log and the names of the project's variables to the AI provider you connected, and may propose buttons such as Add DATABASE_URL, which open the variable form with the key filled in. Nothing runs on its own.
- If Sentry is connected and the deployment has a commit, a Sentry · since this deploy card shows the errors since that commit, with Open the release.
You can't, and why
- Read a deployment's build log on Netlify or DigitalOcean: Netlify publishes no route for it and DigitalOcean only lists the deploy log of the latest version, under App logs. The deployment says “This provider publishes no build log PocketCode can read yet.”Third party
- Keep a Render rollback: Render leaves autodeploy on, so “Autodeploy stays on: the next push to the branch undoes this rollback.”Third party
- Trigger a second deployment while one is in flight for the same project on Free (Redeploy on Netlify, Render, Railway or DigitalOcean, deploying from the editor, the AI assistant or a workflow's Deploy step): it stops with “Upgrade required to trigger another active deploy” and opens the Pro wall.Pro
Rollback on DigitalOcean, in three phases
- 1Roll back on DigitalOcean first asks DigitalOcean: “Asking DigitalOcean whether this rollback is safe…”
- 2You read what DigitalOcean says before anything changes (“This is what DigitalOcean says before anything changes.”) and tap Continue.
- 3The rollback is done, but the app is pinned: it will not accept new deploys until you make this permanent or undo it. Choose Make it permanent or Undo the rollback.
- 4It ends with “Rollback committed. The app accepts deploys again.”, “Undone. The app is back on its previous spec.” or, if DigitalOcean refuses, “DigitalOcean refused this rollback. Nothing has changed.” The whole flow needs Pro.
Logs
Build logs, runtime logs and live logs share one viewer. Reading build and runtime logs is free; the live stream is Pro.
You can
- Search (Search in the log), filter by level (Debug, Info, Warnings, Errors), Pause and Resume, Copy the whole log and Share the log. A LIVE badge shows an open stream. Values that look like a token appear as [hidden], also in what you copy or share.
- Open Filters for Source (Render only: Build, App and Requests, with a text or regular-expression filter that runs on the server), Path (when the lines carry one), Time (Any time, Last 15 min, Last 1 hour, Last 24 hours), Clear filters and Done.
- A logs section reads up to 5 pages and a deployment's Build log up to 20. If there is more it says “The most recent lines. Filter to narrow it down.”; if not, That is everything there is.
- Open Live Logs from a project or from a running deployment (Pro). It goes from Connecting to <Provider>… to Live from <Provider>, counts the lines (N lines streamed), keeps the last 10,000 in memory and stays open only while the screen is. After a minute with no output it says Still nothing after a minute with Reconnect; Stream closed and Stream lost are the other end states.
You can't, and why
- Stream live logs on Netlify or Fly.io: neither publishes a supported live stream.Third party
- Read the HTTP request logs of a Render Hobby workspace: they need a Render Pro workspace and come back empty (“HTTP request logs need a Pro workspace on Render. On Hobby this comes back empty.”)Third party
- Filter live lines by path: they carry no source, so the Path row of the filters is not shown in Live Logs.By design
What each provider streams
| Provider | What Live Logs carries |
|---|---|
| Vercel, Railway | Build output, once something has been deployed (“Nothing deployed yet, so there is nothing to stream”). |
| Cloudflare Pages | Only calls to Pages Functions, once something has been deployed: a static site sends nothing. |
| Render, DigitalOcean | What the running app writes, not the build. The screen says so: “<Provider> streams what the running app writes, not the build”. |
| Netlify, Fly.io | Nothing: neither publishes a supported stream. |
Sections, lists and variables
Every section of a project or an account (deployments, domains, metrics, DNS, variables, users, issues…) is a list that shows only what the provider's official API allows, and says so when it doesn't.
Section lists
A section opens as a list from a project, from an account, from All tools or from a link that names a section.
You can
- Pull to refresh, and Load more for the next page.
- Search (Search this list) when the list has six or more rows, and filter by state with chips (All and the states present) when it has two or more states.
- Read the rows grouped by day (Today, Yesterday) when they carry dates, and the reason when a row failed.
- Per row: open its detail, Copy id, Open in browser and the actions the section declares. Actions of the whole section, such as Add variable or creating a DNS record, sit on the list.
- When a list is empty it says why: “PocketCode does not read this from <Provider> yet” (it exists at the provider and PocketCode doesn't read it), <Provider> does not offer this, or a plain “nothing yet”. The first and the last offer See it in <Provider> to open the provider's panel, and “nothing yet” also offers Retry; “does not offer this” has no button, except for Fly.io's logs.
- Rows marked Open in Database (Supabase, Neon, Turso, Cloudflare D1 and PlanetScale) leave the module and open Database on that project, where the SQL consoles, tables, RLS policies and schema health live.
You can't, and why
- Read what PocketCode doesn't read yet: those sections show “PocketCode does not read this from <Provider> yet”.Not yet
- See more than the provider's official API gives: each section shows only what it allows.Third party
Row detail and CI steps
Tapping a row opens a sheet over the list, so the list keeps its scroll, filter and page.
You can
- Read the metadata the provider sent, Copy all, Copy id, open it in the browser or in the provider's panel (See it in <Provider>), and run the row's actions.
- On a domain that belongs to a connected Cloudflare zone, Your DNS is on Cloudflare jumps to its DNS.
- On GitHub Actions, a job opens with the chips Jobs and Failed at <step>, its Steps, and a button to re-run only the failed jobs.
You can't, and why
—
Environment variables
The variables section of a project (Render, Vercel, Netlify, Railway, Cloudflare Pages and DigitalOcean) shows names, not stored secrets.
You can
- Read the variables split into Secrets · values never leave <Provider> and Plain, and filtered by environment (Production, Preview, Development) with counts when the provider has more than one.
- Copy a name with Copy name.
- Add variable (Key and Value; the value field is masked) and Delete variable. Both are Pro: a confirmation that names the variable, plus your fingerprint if the phone can.
- On DigitalOcean the changes go through the app's spec: Preview the change is free and shows what would happen, Save variable is Pro, and changing a variable triggers a deployment.
- On Cloudflare Pages, tap the eye of a plain-text variable to show its value; secrets never come back.
- When Explain this failure proposes adding a missing variable, the Add variable form opens with the key filled in.
You can't, and why
- See a variable's value on Render, Vercel, Netlify, Railway or DigitalOcean: each row says Hidden by <Provider>, because PocketCode never draws a production secret on screen.By design
- Add or delete variables on Fly.io the same way: it has Secrets (names and a digest) that you set with Pro and that apply when the machine restarts.Third party
Sentry and observability
Sentry, PostHog, Axiom, Healthchecks.io and UptimeRobot appear in Projects under Observability, and what they report feeds Status and the background watch.
You can
- Sentry: Issues, Latest event, Crash-free by release, Cron monitors, Logs and Quota. An issue opens with Events, Users, Release, an Events (24h) chart, its Stack trace (the frames of your code) and Tags, with Open in Sentry and Copy stack.
- The four actions on a Sentry issue (resolve, assign, mute, archive) are reversible, so they are free. They need
event:writeon the token: create it withorg:read,project:read,event:readandalerts:read, and addevent:writeonly if you want to change issues from here. - PostHog: insights, events and daily active users, persons and their timeline, surveys, error tracking and Feature Flags.
- UptimeRobot: monitors, Incidents, Response time, Maintenance windows, Watch a new URL and Silence for a while (a maintenance window of 15 minutes to 2 hours). Healthchecks.io: checks and their flips. Axiom: logs, event volume and query budget.
You can't, and why
- Open the PostHog Feature Flags list on Free: opening it is Pro, and Turn off and Rollout percentage (a 0 to 100 % slider) need Pro too.Pro
- Pause a monitor from the app: UptimeRobot and Healthchecks.io have no reachable control for it. Silence for a while (UptimeRobot) quiets one with a maintenance window.Not yet
Hosting provider by provider
What you can do on each of the seven hosting providers today, taken from what PocketCode implements for each one. Expo EAS is in the second table.
| What you want | Where it works | Where it doesn't |
|---|---|---|
| History of deployments (Deployments) | Vercel, Netlify, Render, Railway, Cloudflare Pages, DigitalOcean. | Fly.io has no deployments: its Activity shows the release in progress and the machines. |
| Redeploy | Netlify, Render, Railway, DigitalOcean. Vercel through its own Redeploy action. | Cloudflare (only Retry this deploy on a failed one) and Fly.io. |
| Roll back (Pro) | Vercel (promotes a previous deployment), Netlify, Render, Railway, Cloudflare Pages, DigitalOcean (three phases for Roll back to this one). | Fly.io. |
| Cancel deploy | Vercel, Netlify, Render, Railway. | Cloudflare, DigitalOcean and Fly.io. |
| Build log of a deployment | Vercel, Render, Cloudflare, Railway. | Netlify publishes no route for it. DigitalOcean's App logs is the deploy log of the latest version only. Fly.io has none. |
| Runtime logs (a section) | Vercel (Runtime logs), Render (Logs, with a source filter), Railway (Logs), Cloudflare (Workers logs). | Netlify and Fly.io (See it in Fly.io) have none; DigitalOcean only through Live Logs. |
| Live Logs (Pro) | Vercel and Railway (build output), Render and DigitalOcean (the running app), Cloudflare Pages (Pages Functions only). | Netlify and Fly.io: no supported stream. |
| Environment variables | Render, Vercel, Netlify, Railway: names plus add and delete (Pro). Cloudflare Pages (production; the form picks the environment). DigitalOcean through the app's spec. | Fly.io has Secrets instead: names and a digest, set with Pro, applied on restart. |
| Domains and DNS | Vercel (the project's and the account's; adding one is Pro), Netlify, Railway (a Railway domain is Pro), Cloudflare (zones, with DNS records), DigitalOcean (domains and DNS records), Fly.io (certificates; checking DNS is free, adding a domain is Pro). | Render has no domains, DNS or TLS section. |
| TLS and certificates | Fly.io (certificates with the days left) and Cloudflare (a zone's certificates). | Vercel, Netlify, Render, Railway and DigitalOcean have no TLS section. |
| Metrics | Vercel (Visitors), Render (CPU, memory, bandwidth), Railway, Fly.io (HTTP responses), Cloudflare (Worker metrics), DigitalOcean (app and droplet). | Netlify has none: its only figure is the day's deploy quota, an estimate. |
| Usage and cost | DigitalOcean: Month-to-date spend and invoices. Vercel: Deployments in the last 24 h. Railway: Usage this cycle, an estimate with no amount. | Render has no usage or cost. Cloudflare's is not read yet. Fly.io has none. |
What else each provider has
| Provider | Also there |
|---|---|
| Vercel | Point alias here, Switch team, pausing (Pro) and resuming a project, and the domains of the project and of the account. |
| Netlify | Locking and unlocking auto-publishing (Pro). |
| Render | Events, the build configuration, editing or scaling an instance (Pro), Suspend (Pro) and Resume. |
| Railway | Approving a pending deployment, and creating a Railway domain (Pro). |
| Fly.io | Machines: events, processes, draining and resuming traffic, restart, stop and start (reversible, free). Secrets, certificates and HTTP responses. |
| Cloudflare | Workers (traffic split and taking a version to 0 %, Pro), KV, R2, D1 (it opens in Database), DNS zones (creating and deleting records is Pro), certificates, WAF, security events and purging the cache (free). |
| DigitalOcean | Droplets (their actions are Pro), Kubernetes clusters, managed databases, volumes, firewalls, uptime checks, invoices and Force rebuild. |
| Expo EAS | A read-only list of your builds through Expo's internal API, best effort. |
The Services tab
The third tab lists your connected services and the catalog of 33. This page only covers what you see in the tab; how to connect each service, the token permissions and the catalog itself are in Platform Services.
You can
- Search all services at once (Search 33 services), by name.
- Read Connected: one row per service with its logo, its name and its number of projects, the failed ones first. Five are visible and N more unfolds the rest.
- Switch the active account with the chips under a row, if you connected more than one account of that service.
- Tap a connected service to Manage connection (Open account when it has an account screen): check it, replace the key or disconnect it.
- Browse Connect, the catalog in eight groups. Tap a service to open its connection sheet over the module; it asks for your fingerprint or screen lock first, if the phone has one.
- On a wide window the list is on the left and Open account and Manage connection are on the right (“Pick a service on the left.”).
You can't, and why
- Connect a second account of the same provider on Free: the first account of each provider is free, the second needs Pro (More than one account per provider).Pro
- Open an account for a service without an account screen: the Open account button is not drawn.Not yet
- Open the connection sheet without confirming it's you, if the phone has a fingerprint or screen lock: the sheet asks first. A phone with neither opens it directly.Device
The eight groups of the catalog
| Group | Services |
|---|---|
| Hosting (8) | Vercel, Netlify, Render, Railway, Fly.io, Cloudflare, DigitalOcean, Expo EAS. |
| Backend (6) | Firebase, Supabase, Appwrite, Convex, PocketBase, Clerk. |
| Data (5) | Neon, Turso, PlanetScale, MongoDB Atlas, Upstash. |
| Observability (5) | Sentry, PostHog, Axiom, Healthchecks.io, UptimeRobot. |
| CI (3) | GitHub, GitLab, Bitbucket. |
| Media (1) | Cloudinary. |
| Payments (2) | Stripe, RevenueCat. |
| Messaging (3) | Resend, OneSignal, Twilio. |
Notifications and background watching
Backend watch lets the phone check your services in the background and notify you, with a widget and a quick-settings tile. It is Pro and it doesn't cover the rest of the notification settings. Open it in Settings → Notifications → Backend Manager (the row has a PRO badge); on Free the screen is a block that opens the wall Let the phone watch it for you.
You can
- Choose how each connected service is watched with Notify, Watch or Off. Watch keeps it in Status without notifications; Off makes no requests. Picking Notify asks for notification permission on Android 13 and later.
- Choose What to notify: switches per kind, in the groups Deploys, CI, Uptime and Errors (each with Send a test) and Also (Domain expiring, Usage above 80%, Provider incidents and Watching stopped). There are 12 kinds in 8 groups. Failures sound as your Android settings say; recoveries arrive quietly.
- Mute single projects with Mute projects: a muted project never notifies but stays in Status. Only services set to Notify are listed.
- Read the cadence card: “Every 15 minutes, and every 45 s for 10 minutes after a deploy from the app.”, “up to 15 min late, no servers in between” and the day's request count.
- A daily pass also checks domains, certificates and quotas. Seven providers have their own incremental query (Render, Cloudflare, DigitalOcean, Sentry, Healthchecks.io, GitHub and GitLab); for the rest the watch compares the last known state.
You can't, and why
- Get real-time notifications: it is polling on the phone, not push. Android's minimum for a periodic job is 15 minutes, so a notification can arrive up to 15 minutes late, and no PocketCode server polls for you.By design
- Turn off Watching stopped (“A token expired · always on”): you always get told when a token stops working.By design
The three pages of the screen
Services
The cadence card, the request count and Notify / Watch / Off for each service. If a 401 switched watching off, it says “Off: this token no longer works. Reconnect the service.”
What to notify
The 12 kinds in 8 groups, each kind with its own switch; Send a test sits on the groups Deploys, CI, Uptime and Errors.
Mute projects
Mute and Unmute for single projects; a muted one reads Muted · still in Status.
Widgets, tile and shortcuts
- The widget Backend status shows up to five projects, the broken ones first. The widget Backend project shows one project you pick when you add it: its state and last deploy, and a Redeploy button (Re-run for a CI repository) that opens the app and never deploys.
- The Backend tile in quick settings says All good or how many projects are broken.
- Without Pro, widgets and tile say Backend watch is Pro; the widgets add “Open the app to unlock it.”
- The launcher shortcuts Status and Redeploy need no plan: they only open screens.
Deploy from the editor
A sheet over the app that doesn't need the module open. Open it with Deploy this project in the file explorer's action menu, or Deploy Project in the editor's menu and command palette (only when the project is on disk). A short wizard saves the destination in .pocketcode/deploy.json inside the project, with no token in it, and then a summary offers Deploy.
The wizard, step by step
- 1Pick the Provider. Only providers you connected that have a deployment history are listed (Render, Vercel, Netlify, Railway, Cloudflare and DigitalOcean). With none connected the sheet says so and offers Open the Backend Manager.
- 2Pick the Site or project from your account.
- 3Pick How: Upload the output folder (ZIP), with its Output folder (
dist,build,out,public,_siteandwwware suggested only if they hold anindex.html), or Commit and push, which stages everything, commits if there are changes and pushes the current branch (set a Commit message; the default is Deploy from PocketCode). - 4Tap Save destination.
- 5On the summary tap Deploy. A confirmation says “This uploads … to production”, and asks for your fingerprint or screen lock if the phone can.
- 6At the end, Watch the deploy opens the deployment or the Deployments list and, if the remote is GitHub and connected, Watch CI opens its runs. If Sentry is connected, a Sentry release is also created for the commit.
What else to know
- Change destination returns to the wizard. While the provider limits deploys the sheet says The provider limits deploys: wait N s.
- The ZIP leaves out
.git,.pocketcode,node_modules, symbolic links, hidden files and key files (id_rsa,*.jks,local.properties…); only.well-knownis uploaded, and the sheet says how many files were left out. deploy.jsonhas no tokens and can be edited by hand; if it can't be read, the sheet asks you to set the destination again.
You can't, and why
- Upload a ZIP to anything but Netlify: Vercel and Cloudflare Pages appear disabled with Not read yet by PocketCode.Not yet
- Use Commit and push without a git remote (Needs a git remote in this project) or with merge conflicts (“Resolve the merge conflicts before deploying.”).By design
- Deploy an empty output folder: it would replace your site with nothing.By design
- Upload more than 200 MB or 20,000 files in the ZIP.By design
- Deploy to Fly.io: it has no deployment history, so it is not offered.Third party
- Deploy while the project already has a deployment in flight on Free: the sheet shows the error “Upgrade required to trigger another active deploy” and a See Pro button that opens the paywall.Pro
Free and Pro
Reading is free on every plan, with two exceptions: live logs and the PostHog feature flag list. Pro adds writing, background watching, the AI explanations, several accounts per provider and no cap on deployments in flight. The module has eight Pro gates.
| Feature | Free | Pro |
|---|---|---|
| Read: lists, status, build and runtime logs, metrics, usage, domains and DNS, variable names, Deployments, Timeline | Included | Included |
Connect the 33 services (MULTI_ACCOUNT) | All 33, one account per provider | All 33, several accounts per provider |
Deployments in flight per project (UNLIMITED_DEPLOYS) | 1 | Unlimited |
Roll back, Roll back to this one and Promote to production (DEPLOY_ROLLBACK) | Not included: the button shows and the Pro wall appears after you confirm | Included |
Live Logs (LIVE_LOGS) | Not included: the Pro wall opens when you tap it | Included (Vercel, Render, Railway, Cloudflare Pages, DigitalOcean) |
Destructive changes (BACKEND_WRITE): Suspend, add or delete variables, DNS records, domains, scaling, droplet actions, Fly.io secrets, Neon Suspend compute and Start compute, Convex Run mutation, the Cloudflare Worker traffic split and Take a version to 0% | Not included | Included, each with a confirmation and your fingerprint if the phone has one |
PostHog Feature Flags (FEATURE_FLAGS) | Not included, not even to read | Included; turning one off or changing its rollout also needs BACKEND_WRITE |
Background watching (BACKEND_WATCH): checks, notifications, widgets and the tile | Not included | Included |
AI (BACKEND_AI): Why it failed, Explain this failure, Summarize today | Not included | Included; it uses your own AI provider |
| Free on every plan: Redeploy (under the cap), Re-run, Cancel deploy, Resume, Retry this deploy, purging Cloudflare's cache, Fly.io machine actions, Sentry issue actions | Included | Included |
Privacy and security
What stays on the phone and what leaves it.
- Each service is connected with your own token. It is stored encrypted with the Android Keystore (AES-256-GCM), in a separate file for each PocketCode account, so another account on the same phone can't read it.
- There is no PocketCode server in between: every call goes from your phone to the provider's official API, and the provider bills what you use under its own plan. A PocketBase or a self-hosted Appwrite goes to the server whose address you type. Platform incidents come from the providers' public status pages, with no token.
- The window blocks screenshots and screen recording, and so do the sheets that hold secrets. The connection sheet is also protected against tapjacking.
- The value of a variable is never shown on Render, Vercel, Netlify, Railway or DigitalOcean, and the value field of a new variable is masked while you type.
- Lines that look like a token appear as [hidden] in the log viewer, so also in what you copy or share.
- Every change asks first. Reversible actions ask once, naming the resource; destructive ones also ask for your fingerprint or screen lock if the phone has one; irreversible ones are never offered.
- Explain this failure sends an extract of the build log (with the values it knows masked), the project and deployment names and the names of the project's variables to the AI provider you connected. Summarize today sends the projects that need attention. Neither runs anything by itself.
- Disconnecting a service removes the key from the phone only: it keeps working at the provider until you revoke it there.
- Watching runs on the phone: no PocketCode server polls your providers or receives your tokens. The log of watch events and the log of what you changed today stay on the phone.
Limitations
What the module can't do, grouped by reason. Each item says why.
- Everything needs a connection and your own token. Offline you see the last data loaded (Offline · showing data from …) or No connection.Needs network
- Some providers' request limits (Cloudflare's, for one) are shared with their own dashboard: PocketCode loads only what you open, refreshes only the visible providers and waits when a limit is reached (Request limit reached).Third party
- Netlify allows three deployments a minute, so Redeploy is disabled for 20 seconds after you use it.Third party
- Netlify has no build-log or live-log API, and Fly.io has no supported log API, deployment history or rollback: those parts are missing for them, not hidden.Third party
- Live Logs on Cloudflare Pages only carry calls to Pages Functions, and the HTTP request logs of Render need a Render Pro workspace.Third party
- Incidents only come from the 14 providers with a known public status page.Third party
- Irreversible actions (deleting resources,
DROP, restoring backups) are never offered, and no advanced mode exists to enable them.By design - Variable values are never shown on Render, Vercel, Netlify, Railway or DigitalOcean.By design
- Watching is polling on the phone every 15 minutes, with a 45-second window for 10 minutes after a deploy you make from the app and a daily pass for domains, certificates and quotas.By design
- A ZIP deployment from the editor only works on Netlify; on Vercel and Cloudflare Pages it shows Not read yet by PocketCode.Not yet
- Links, widgets and the launcher's Redeploy only work for 13 providers, and there is no control to pause a monitor.Not yet
- If the phone has no fingerprint or screen lock, a risky action still goes ahead, with its confirmation.Device
- Destructive changes, rolling back, live logs, feature flags, the AI explanations, background watching, several accounts per provider and more than one deployment in flight per project are Pro. Reversible actions are free.Pro
Troubleshooting
The titles are the messages the app shows. <Provider> stands for the name of the provider, and … for the folder name the real message includes.
<Provider> rejected the token
The token expired or was revoked. Open Services, tap the service and replace the key, or tap Reconnect on the card.
This token can't read this
The provider says the token lacks a permission. Create a new token with the permissions the connection sheet lists and reconnect with it.
No connection
PocketCode couldn't reach the provider. Check your connection and try again. While it shows the last data it loaded, a strip says Offline · showing data from ….
Request limit reached
PocketCode waits so the limit you share with the provider's dashboard can refill. The button counts down (Retry in N s), and in Projects a provider in backoff shows Wait N s.
Not in your <Provider> plan
The provider only offers this on a paid plan. There is nothing to fix in PocketCode.
Included with Pro
The section is part of PocketCode Pro. Today that is the PostHog Feature Flags; See Pro opens the paywall.
“Upgrade required to trigger another active deploy”
On Free, a project can have one deployment in flight. Wait for it to finish, cancel it, or go Pro.
“No earlier successful deploy to go back to”
The header Roll back found no earlier deployment that worked. Open the deployment you want and use Roll back to this one.
“This provider publishes no build log PocketCode can read yet.”
Netlify and DigitalOcean have no build log per deployment. On DigitalOcean, App logs holds the deploy log of the latest version.
“Nothing deployed yet, so there is nothing to stream”
Live Logs on Vercel, Cloudflare and Railway need at least one deployment to stream.
Still nothing after a minute
The stream is connected but silent. Tap Reconnect. On Cloudflare a static site sends nothing, and on Render the app has to write lines.
“Off: this token no longer works. Reconnect the service.”
Watching stopped for that service because its token was rejected. Reconnect it from Services. The notification Watching stopped can't be turned off.
“PocketCode does not read this from <Provider> yet”
The data exists at the provider but PocketCode has no screen for it yet. See it in <Provider> opens the provider's panel.
“Autodeploy stays on: the next push to the branch undoes this rollback.”
Render keeps autodeploy on after a rollback. Switch it off in Render if you want the rollback to last.
“HTTP request logs need a Pro workspace on Render. On Hobby this comes back empty.”
Request logs are a Render Pro workspace feature. Build and app logs are not affected.
“The output folder «…» doesn't exist or is outside the project.”
Fix Output folder (it is relative to the project root, and “.” is the whole project) or edit .pocketcode/deploy.json.
“The output folder «…» is empty: deploying it would replace your site with nothing.”
PocketCode refuses to deploy an empty folder. Build your site first, or pick the folder that holds it.
“The output folder is too big to upload from the phone”
The ZIP limit is 200 MB and 20,000 files, counted without compressing. The message gives the file count and size at the point where PocketCode stopped counting. Pick a smaller folder.
“This project has no git remote to push to.”
Commit and push needs a git remote in the project, and it stops on merge conflicts (“Resolve the merge conflicts before deploying.”).
“The file .pocketcode/deploy.json can't be read”
The file is invalid. Set the destination again from Deploy this project; it holds no tokens, so you can also fix it by hand.
No AI provider connected
Summarize today and Explain this failure ask the AI provider you connected in PocketCode; connect one in the AI settings first. This message is always in English.
“PocketCode has no screen for this provider yet.”
You tapped the name of a provider that has no account screen in Backend Manager.
Next steps