getcatalystiq

Agent Plane — AI skill for Claude Code

AI community

A multi-tenant platform for running AI agents in isolated Vercel Sandboxes, exposed via a REST API.

How to install Agent Plane

This entry records only its repository, not the path inside it, so there is no exact command to give. Open getcatalystiq/agent-plane and copy the folder into ~/.claude/skills/, or the file into ~/.claude/agents/.

What Agent Plane does

Alternatives in AI

Coding Agent Template — Multi-agent AI coding platform powered by Vercel Sandbox and AI Gateway 1.8k ★
CliRelay — Self-hosted AI gateway for coding CLIs — one OpenAI/Claude/Gemini/Codex-compatible endpoint, with a multi-tena 1k ★
My Free Code — Open-source multi-provider AI gateway for Claude Code and other coding agents, with model routing, streaming 633 ★

README

AgentPlane

A multi-tenant platform for running AI agents in isolated [Vercel Sandboxes](https://vercel.com/docs/sandbox), exposed via a REST API. Supports any model via [Vercel AI Gateway](https://vercel.com/docs/ai-gateway) with dual runner architecture ([Claude Agent SDK](https://docs.anthropic.com/en/docs/claude-code/sdk) + [Vercel AI SDK](https://sdk.vercel.ai)).

**AI Agents as an API** — one POST request to launch an agent with skills, MCP connectors, streaming, and full observability. Built for platforms.

Features

Any model, any provider — supports all models via Vercel AI Gateway; dual runner: Claude Agent SDK for Anthropic models, Vercel AI SDK for everything else
Isolated sandboxes — every execution spins up a Vercel Sandbox from a pre-built SDK snapshot (~3s cold start) with its own filesystem, network policy, and resource limits
Skills & plugins — inject custom skills and tenant-scoped marketplace plugins into agents before execution
Scheduled executions — configure agents to run on a schedule (hourly, daily, weekdays, weekly) with timezone-aware execution
Multi-tenant — row-level security, per-tenant API keys, budget controls, and rate limiting
MCP connectors — connect agents to external tools via Composio (GitHub, Slack, Firecrawl) or tenant-scoped custom OAuth 2.1 MCP servers
A2A protocol — expose agents to external A2A-compliant clients (LangGraph, CrewAI, Semantic Kernel, etc.) via Agent Cards and JSON-RPC; streaming SSE, idempotency, budget clamping
Full observability — every session message stores a transcript, token usage, cost, duration, and trigger source (API/schedule/playground/chat/A2A/webhook)
Playground — test agents interactively from the admin dashboard with real-time event streaming
SoulSpec v0.5 identity — full SoulSpec compliance: SOUL.md (Personality/Tone/Principles), IDENTITY.md (Name/Role/Creature/Emoji/Vibe), STYLE.md, AGENTS.md, HEAR