Your own AI developer. Tell it what you need — it writes the code, opens pull requests, and ships it.
No coding experience required. You describe what you want in plain English, and your agent builds it.
Website | Docs | Blog | API Reference
Option A — one-line installer (auto-detects everything):
curl -fsSL https://raw.githubusercontent.com/CorvidLabs/corvid-agent/main/scripts/install.sh | bashOption B — clone and setup:
git clone https://github.com/CorvidLabs/corvid-agent.git && cd corvid-agent
bun run setup # guided wizard: detects AI provider, creates .env, installs deps
bun run dev # starts server + dashboard at http://localhost:3000Option C — Docker (no Bun/Node required):
git clone https://github.com/CorvidLabs/corvid-agent.git && cd corvid-agent
cp .env.example .env # edit with your API keys (or leave defaults for local Ollama)
docker compose up -d # builds and starts at http://localhost:3000Add corvid-agent tools to your AI editor (Claude Code, Cursor, Copilot):
corvid-agent init --mcp # configures MCP server + copies Agent SkillsFull setup guide → | MCP setup → | VibeKit integration →
An open-source AI agent platform that writes code, opens pull requests, and ships software. It combines LLM-powered coding with on-chain identity (Algorand/AlgoChat), multi-agent orchestration, and integrations with Discord, Telegram, Slack, and GitHub.
You describe what you want in plain English. Your agent designs, codes, tests, and deploys it.
- "Build me a weather dashboard" → it built this
- "Make a movie browser for classic films" → it built this
- "I need an earthquake tracker" → it built this
- "Create a poetry explorer" → it built this
- "Build a pixel art editor" → it built this
Every app above was designed, coded, tested, and deployed by corvid-agent — zero human-written application code. See all apps →
| What you get | |
|---|---|
| Creators — have ideas but don't code | Describe what you want in plain English. It designs, codes, and deploys. Get started → |
| Developers — write code and want help | Automated PR reviews, CI fixes, test generation, issue triage. Use cases → |
| Teams — need AI to handle dev work | Agents review code, write features, ship PRs. Business guide → |
| Enterprise — need security, compliance, scale | Multi-tenant, RBAC, audit trails, Docker/K8s. Enterprise guide → |
| Channel | What you need |
|---|---|
| Web dashboard | Nothing — included at http://localhost:3000 |
| Terminal | corvid-agent (interactive CLI) |
| Discord / Telegram / Slack | Add a bot token to .env |
| Your AI editor | corvid-agent init --mcp (Claude Code, Cursor, Copilot, etc.) |
corvid-agent handles dev orchestration. VibeKit handles blockchain operations. Together they give you a complete Algorand development stack:
corvid-agent init --mcp # add corvid-agent MCP tools
vibekit init # add blockchain MCP tools (deploy, assets, indexer)Your AI editor gets 50 corvid-agent tools (code, GitHub, scheduling, agents) plus 42 VibeKit tools (contract deploy, ASA management, transaction signing) — all working side by side.
Skills are markdown files in .skills/ or skills/ that teach AI assistants how to use corvid-agent. Each skill has a short description for discovery and a full body loaded on demand:
skills/
coding/SKILL.md # File operations, shell commands
github/SKILL.md # PRs, issues, reviews
smart-contracts/SKILL.md # VibeKit + Algorand contract tools
scheduling/SKILL.md # Cron-based task automation
...30 skills total
corvid-agent init --mcp copies skills to your editor automatically. Skill list →
Bun + Angular 21 + SQLite + Claude Agent SDK + Algorand (on-chain identity). 50 MCP tools, ~380 API endpoints, 10,500+ test files, 193 E2E spec files.
Architecture → | Security → | Deployment →
Open source because AI agents should be owned by the people who run them.
- Good first issues
- CONTRIBUTING.md — setup takes ~2 minutes
- Discussions