Editing connection access
Connection settings explain missing choices and ask for your password when you save.
Changelog
What's new.
Every shipped change, newest first.
Connector access and security checks
Choose which apps your assistant can work on, and see what to do when a security check stops a publish.
A calmer dashboard
Shared controls across the dashboard, docs and pricing page, grouped logs, and tables that adapt to narrow screens.
Crash reports name the likely cause
Crash reports recognise common causes, keep the memory advice, and give you a prompt for your own AI assistant.
Two Free apps, a smaller badge and steadier rollbacks
Free now covers two apps and says what to do when a start does not fit. Free apps show a small Percher icon. Rollback, previews and the 1.0.3 tools are more reliable.
Percher 1.0: one version, a clear licence boundary
The six npm packages establish the 1.0 licence boundary. The 1.0.1 patch clarifies shared data during previews; shared and toml remain at 1.0.0.
Preview URLs where you look for them, and a faster Publishes tab
Every active preview is now one click away, and the publish history loads instantly.
The capsule template — a full-stack starter in one command
Scaffold an app with a database, sign-in, and realtime already wired up, then publish. (The managed-AI part is currently switched off.)
Redirects and response headers in percher.toml
Declare redirects and custom response headers in your toml, applied before a request reaches your app.
Deploy hooks
A secret URL that rebuilds your app — trigger it from a CMS, a cron job, or CI.
Deploy a GitHub repo without a percher.toml
Connect a repo with no config — Percher detects the framework and generates one just for that deploy.
Scoped API tokens
Give CI and coding agents a token scoped to deploy, env, or read-only — optionally bound to one app.
Percher is now a one-click connector for Claude.ai
Add `https://mcp.percher.app/mcp` as a custom connector in Claude.ai, Cursor, or Windsurf. OAuth login, no token to paste, and your assistant can operate your apps and deploy connected repos.
Separate RSS feeds for the blog and changelog
The blog and changelog now each have their own RSS feed, so you can follow just the one you want.
See what changed, and take your history with you
New `percher diff` shows what changed between deploys, and `percher export --git` hands you your whole history as a git bundle.
Publish without an account
Tell your AI assistant to publish and your app goes live in seconds, no signup. Claim it to a real account whenever you want to keep it.
Read-only inspect links
Share a link to your app's live status and recent deploys — no login, nothing sensitive, revocable any time.
Automatic security scanning
Percher now scans every app image for known vulnerabilities, surfaces them in your Security tab, and emails you only when something critical needs a fix.
Verified email at signup
New accounts confirm their email and accept the terms before going active, with a one-click resend if the link expires.
Export your app and data anytime
Download your whole app — source, config, env keys, and database — as a single archive. No lock-in.
Health checks that wait for slow starts
Apps that run migrations or warm a cache on boot can now extend the health-check window in percher.toml.
Invite emails for collaborators
When you invite someone to an app, they now get an email with the role, what it can do, and a link straight to accept.
Pool sizing update
Starter gets a full 1.0 vCPU pool, disk pools across paid tiers were resized to match the VPS steps we actually run on, and the app cap was raised on every tier.
Account deletion shows you what's in the way
Settings → Danger zone now lists everything blocking account deletion before the button is enabled — owned apps, collaborators, in-flight deploys, pending transfers.
Resource pools per account
Each paid tier now sells a per-account RAM, vCPU, and disk pool that you distribute freely across your apps.
Static site hosting
Static sites run on a slim, fast Caddy-based container — quick cold starts and small images.
Faster builds for Node and Bun apps
Node and Bun apps build through a purpose-built pipeline — cold builds in ~15–25s, warm redeploys in ~10–15s.
Template gallery
Browse starter templates at percher.app/new and start one with a single command.
Filter logs by level and date
Narrow runtime logs by severity and date range, with shareable URLs that preserve the filter.
A smoother deploy flow for AI agents
percher_publish returns immediately and hands the assistant a wait-for-deploy step, so agents stay responsive during long builds.
GitHub App, PR previews, and status checks
Install the Percher GitHub App once — every push deploys, pull requests get preview URLs, and deploy status posts back as a commit check.
Graceful zero-downtime deploys
A new version goes live only after it passes a health check; the running version keeps serving until the swap completes.
Scale across multiple instances
Apps can run multiple container instances, with optional CPU-based autoscaling configured in percher.toml.
Backups and restore
Every app's database is backed up nightly and replicated off-site; restore from the dashboard or with percher restore.
Signed deploy webhooks
Subscribe to deploy events with HMAC-SHA256-signed webhooks and wire Percher into your own automation.
Password-protect any app
Add password = true to percher.toml and visitors get a branded login page before they reach your site.
Percher is live
Say "publish my app" to your AI assistant — or run bunx percher publish — and your app is live at name.percher.run with SSL.
AI-powered crash diagnostics
When an app crashes, Percher detects it, explains common causes and emails you.
Prometheus metrics and an OpenAPI spec
The Percher API ships a full OpenAPI spec, and the platform exposes operational metrics in Prometheus format.