Open source infrastructure for tool-using AI systems.

We build systems around a simple idea: keep model context small, keep runtime behavior explicit, and make protocol surfaces practical to deploy.

• Docs: https://zerocontextprotocol.vercel.app/
• Python SDK: https://pypi.org/project/zero-context-protocol-sdk/
• Protocol repo: https://github.com/FishCodeTech/zero-context-protocol
• SDK repo: https://github.com/FishCodeTech/zero-context-protocol-python

zero-context-protocol is the public docs and architecture repository for ZCP.

• Protocol design
• Architecture notes
• Benchmarks and methodology
• MCP compatibility notes
• Deployed docs site

zero-context-protocol-python is the production Python SDK for ZCP.

• Native ZCP runtime
• MCP-compatible surfaces
• HTTP, WebSocket, and stdio transports
• OAuth and task support
• Published package: zero-context-protocol-sdk

Install:

pip install zero-context-protocol-sdk

ctf-agent-benchmark is a benchmarking project for evaluating AI agents in security and tool-use workflows.

• CTF-oriented evaluation tasks
• Agent capability benchmarking
• Security workflow experiments

Most tool-calling stacks push too much machine metadata into model context. We care about the other direction:

• smaller visible tool surfaces
• tighter runtime validation
• better context reuse
• cleaner protocol boundaries

That is the design center behind ZCP.

• Docs: https://zerocontextprotocol.vercel.app/
• PyPI: https://pypi.org/project/zero-context-protocol-sdk/
• Benchmarks: https://zerocontextprotocol.vercel.app/benchmarks
• Architecture report: https://zerocontextprotocol.vercel.app/architecture