Forkit
@zientesit
Persistent coordination infrastructure for multi-agent AI systems, exposed as a single MCP endpoint.
Overview
What is Forkit?
Forkit is an MCP server that provides shared, persistent infrastructure for agent swarms. It enables agents to coordinate through a shared task graph, claim tasks atomically, resolve dependency graphs automatically, hand off work via long-poll, and execute code in a real V8 sandbox. Built on Cloudflare Workers and D1, it is designed for multi-agent systems needing race-condition-safe task management and dramatic token reduction.
How to use Forkit?
Integrate Forkit as an MCP server by connecting your MCP client to its endpoint. Agents use the provided tools to create tasks, claim them atomically, resolve dependencies, execute async JavaScript via the execute_code tool, and receive task handoffs via long-poll. No specific configuration keys or installation steps beyond standard MCP client setup are documented.
Key features of Forkit
- Shared task graph with atomic, race-condition-safe claiming
- Automatic dependency graph resolution for task ordering
- Long-poll handoff with ~250ms wake latency
- Execute_code tool runs async JS in a real V8 sandbox
- Collapses multi-step workflows from ~150K tokens to ~1K
- Built on Cloudflare Workers and D1 with usage-based pricing
Use cases of Forkit
- Coordinating a swarm of AI agents that need to share and claim tasks
- Reducing token overhead in multi-step agent workflows by offloading code execution
- Building persistent, dependency-aware task pipelines that survive agent restarts
- Enabling real-time task handoffs between agents with low-latency polling
FAQ from Forkit
How does Forkit handle task claiming?
Forkit uses atomic, race-condition-safe claiming to ensure only one agent can claim a task at a time.
What is Forkit's pricing model?
The first 50 tasks are free. After that, each task costs $0.01 USDC, paid via x402.
What runtime does Forkit require?
Forkit runs on Cloudflare Workers and uses D1 as its database. No local installation is needed.
How does the execute_code tool work?
It runs asynchronous JavaScript in a real V8 sandbox, allowing agents to collapse multi-step workflows from ~150K tokens to ~1K.
What transport does Forkit use?
Forkit operates over a single MCP endpoint and uses long-poll for task handoff with approximately 250ms wake latency.
