ThinkNEO MCP Server is part of the ThinkNEO platform and is positioned as an open-source MCP server for enterprise AI governance. It provides 72 tools covering governance, guardrails, FinOps, intelligent routing, observability, compliance, outcome verification, and other areas, and includes a built-in production defense layer called ThinkShield.
Key capabilities
- ThinkShield defense layer: includes a detection engine, 5 rule packs, and runtime middleware, reportedly with 145 tests and p99 under 1ms.
- 72 MCP tools: by category, these include governance (6), guardrails (3), FinOps (4), intelligent routing (4), trust scoring (2), registry (5), bridging (4), observability (5), business value (6), A2A control (4), optimization (1), outcome verification (4), policy engine (4), benchmarking (3), compliance (2), Agent SLA (4), audit export (3), caching (3), security (5), tokens (1), memory (2), scheduling (1), and alerting (1). For the full tool reference, see docs/quickstart.md.
- 24 A2A skills: provide bidirectional bridging between MCP and A2A, supporting Google's Agent-to-Agent protocol (v0.3.0, Linux Foundation).
- Protocol compliance: fully implements Model Context Protocol 2024-11-05 and is forward-compatible with MCP 2025-03-26.
How to use It can be run via Docker, or connected from Claude Desktop, Cursor, or any MCP client. The free tier offers 500 calls per month, automatically provisions an API key, and all 72 tools are available.
Use cases Suitable for scenarios requiring governance, security protection, observability, cost management (FinOps), and compliance control during AI agent runtime, especially for enterprise AI systems using the MCP or A2A protocols.
Limitations and boundaries
- This is not the complete ThinkNEO platform. Governance orchestration, encrypted audit chains, tenant management, and enterprise integrations are proprietary features that run on thinkneo.ai.
- It is not a standalone security product. ThinkShield is a defense component within a larger governance platform.
- The project is licensed under Apache-2.0 and is free touse, fork, and contribute to.
- For security vulnerability reports, see SECURITY.md, and for contribution guidelines, see CONTRIBUTING.md.
- Developed by ThinkNEO AI Technology Co., Ltd. (Hong Kong CR No. 2296774); the team also built the ThinkNEO Enterprise AI Control Plane, the AIRGP protocol, and the A2ASTC conformance suite. Partners include NVIDIA Inception, Anthropic Partner, and Linux Foundation A2A.
