The workspace for everything your AI agents create.
staqly is a local-first, self-hostable workspace where agents store what they generate — and where you approve, reject, edit and automate it. REST, MCP and signed webhooks included.
- 1
- Go process
- 1
- SQLite file
- 0
- API keys locally
❯ claude mcp add --transport http staqly http://localhost:3333/mcp
❯ “Add these 4 avatars to Instagram Avatar”
⏺ staqly · create_item ×4 → status pending
⏺ staqly · await_decision …
- Avatar post #1Golden hour on the rooftop ☀️ approved
- Avatar post #2Neon nights in the city 🌃pending
- Avatar post #3Slow morning, good coffee ☕pending
Why staqly
Agents make things. You stay in control.
Everything an agent produces lands in a collection with full history. Nothing ships until a human — or a rule you wrote — says so.
One Inbox for every agent
Images, video, text, links and JSON from Claude Code, Codex, OpenCode, n8n and scripts — reviewed in one place, mostly from the keyboard.
Learn moreHuman approval gates
Agents propose, wait for your decision and act only when you approve — with edits, expiry and a reported result.
Learn moreRules engine
Auto-approve confident output, reject noisy sources, expire what nobody reviews and escalate what is urgent.
Learn moreWatchers
Follow RSS feeds and web pages — Hacker News, Reddit, GitHub, arXiv, YouTube and more — as pending items.
Learn moreDecide from anywhere
Approve or reject from Telegram, Slack, Discord, ntfy or email — and every decision reaches your automations as a signed webhook.
Learn moreLocal-first
One Go process, one SQLite file, one uploads folder. Keyless on your machine; hosted mode when you need it.
Learn moreFive minutes
From zero to your first review.
One command installs it, one starts it: a single binary with the API, MCP and dashboard (or Docker, if you prefer). No account, no API key: on your own machine staqly is keyless.
# 1. install and run it
$ curl -fsSL https://raw.githubusercontent.com/staqly/staqly/main/scripts/install.sh | sh
$ staqly serve
# 2. create a collection and an item
$ curl -X POST localhost:3333/v1/collections -d '{"name":"Drafts"}' \
-H 'Content-Type: application/json'
$ curl -X POST localhost:3333/v1/collections/Drafts/items \
-H 'Content-Type: application/json' -d '{"content":"Hello"}'
# 3. connect your agent
$ claude mcp add --transport http staqly http://localhost:3333/mcpExplore the docs
Connect agents
Local mode — the default
One workspace on your machine. The API answers only this computer, without keys. Phones join with a QR code.
Cloud mode — when you need it
Many workspaces, scoped API keys, sessions and recovery keys, Free / Indie / Team plans. Same app, one environment variable.
Review anywhere. The dashboard is built for phones and tablets, with swipe-to-review.