telos-mcp

telos-mcp

Exposes TELOS governance primitives—action scoring, receipt verification, Purpose Anchor inspection, audit-chain queries, and CCRS counterfactual replay—as MCP tools, resources, and prompts for any MCP-compatible client.

Category
访问服务器

README

telos-mcp

Model Context Protocol server for TELOS AI Labs governance.

Exposes TELOS governance primitives -- action scoring, unsigned integrity-hash receipt verification, Purpose Anchor inspection, audit-chain queries, and CCRS counterfactual replay -- as MCP tools, resources, and prompts. Any MCP-compatible client (Claude Desktop, Claude Code, Cursor, Cline, etc.) can call them as native tools.

v0.1 status. This release ships the protocol surface with stubbed engine calls so clients can exercise every tool, resource, and prompt end-to-end without an installed engine. Scores are synthetic and receipts are unsigned integrity-hash receipts (signature: null, signing_status: "unsigned_integrity_hash"): a key-free SHA-256 over the receipt payload, not a digital signature. v0.2 wires the live telos-gov engine behind the same receipt shape, and a real signature fills the existing signature field then, with no schema break.


What this server exposes

Tools (LLM-callable)

Tool Purpose
telos_score(action_name, action_params, agent_id) Score a proposed action against the active Purpose Anchor. Returns verdict (EXECUTE / CLARIFY / ESCALATE) plus per-dimension scores across purpose, scope, boundary, tool, chain, codebase_fidelity, plus an unsigned integrity-hash receipt envelope.
telos_verify(receipt_json) Verify an unsigned integrity-hash receipt offline (key-free; not a signature check). Returns {match, recomputed_integrity_hash, signing_status, reason, ...}.
telos_get_pa(agent_id) Return the active Purpose Anchor: purpose, hard boundaries, declared scope, allowed tools, centroid dimensions.
telos_audit(start_iso, end_iso) Return audit-chain entries within a time window (capped at 100).
telos_replay(receipt_id, alt_config_path) Queue a CCRS counterfactual replay against an alternate config; returns a job_id.

Resources (LLM-readable)

URI template Returns
telos://wiki/{path} A TELOS governance wiki page by relative path.
telos://centroids/{dimension} Compiled centroid metadata for one of purpose, scope, boundary, tool, chain, codebase_fidelity.
telos://audit/{date} The audit-chain segment for a given YYYY-MM-DD (UTC).

Prompts (user-invokable templates)

Prompt Purpose
governance_review(action) Have the model review a proposed action for governance compliance.
audit_walkthrough(start_date, end_date) Walk through the audit chain in a date window.

Install

pip install telos-mcp

The live engine (telos-gov) lands in v0.2, after the clean-facade republish. v0.1 ships no installable engine extra; the stub is the only supported engine. (telos-mcp[engine] returns in v0.2 pinned to the republished clean facade.)

The package installs a telos-mcp console script that starts the server over stdio. You can also invoke it via python -m telos_mcp.server.

Requires Python 3.10+.


Client configuration

Each client below assumes telos-mcp is installed in a Python environment on PATH. If you prefer not to install globally, use uvx telos-mcp or substitute the absolute path to the telos-mcp script in the snippets.

Claude Desktop

Edit ~/Library/Application Support/Claude/claude_desktop_config.json (macOS) or %APPDATA%\Claude\claude_desktop_config.json (Windows). Add the telos-governance entry under mcpServers:

{
  "mcpServers": {
    "telos-governance": {
      "command": "telos-mcp",
      "args": [],
      "env": {
        "TELOS_AGENT_ID": "claude-desktop-default"
      }
    }
  }
}

Restart Claude Desktop. The five telos_* tools should appear in the MCP tool tray.

Claude Code

Use the claude mcp add CLI:

claude mcp add telos-governance -- telos-mcp

Or edit ~/.config/claude-code/mcp_servers.json directly:

{
  "telos-governance": {
    "command": "telos-mcp",
    "args": []
  }
}

Cursor

Edit ~/.cursor/mcp.json (global) or .cursor/mcp.json (per-project):

{
  "mcpServers": {
    "telos-governance": {
      "command": "telos-mcp",
      "args": [],
      "env": {
        "TELOS_AGENT_ID": "cursor-default"
      }
    }
  }
}

Cline (VS Code extension)

Open Cline -> MCP Servers -> Edit Configuration. Add:

{
  "mcpServers": {
    "telos-governance": {
      "command": "telos-mcp",
      "args": [],
      "transportType": "stdio"
    }
  }
}

Verifying the install

From any client, ask:

Use the telos_get_pa tool with agent_id="test" and show the Purpose Anchor.

You should see a stub PA back containing "stub": true and the six canonical centroid dimensions.


Development

git clone https://github.com/TELOS-Labs-AI/telos-mcp
cd telos-mcp
python -m venv .venv && source .venv/bin/activate
pip install -e ".[dev]"

# Run the server over stdio (what MCP clients spawn):
python -m telos_mcp.server

# Or use the MCP Inspector for interactive exploration:
mcp dev telos_mcp/server.py

Roadmap

  • v0.1 (this release) -- protocol surface complete, engine stubbed, registry-submission-ready manifest.
  • v0.2 -- swap stubs for live telos-gov calls (after the clean-facade republish). Receipts gain a real signature after the signing ceremony, filling the existing signature field with no schema break. telos-gov becomes a hard dependency.
  • v0.3 -- streaming audit subscriptions, per-tenant PA selection via TELOS_TENANT env, optional HTTP transport.

License

Apache-2.0. See LICENSE.

Links

  • TELOS AI Labs: https://telos-labs.ai
  • Source: https://github.com/TELOS-Labs-AI/telos-mcp
  • Issues: https://github.com/TELOS-Labs-AI/telos-mcp/issues

推荐服务器

Baidu Map

Baidu Map

百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。

官方
精选
JavaScript
Playwright MCP Server

Playwright MCP Server

一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。

官方
精选
TypeScript
Magic Component Platform (MCP)

Magic Component Platform (MCP)

一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。

官方
精选
本地
TypeScript
Audiense Insights MCP Server

Audiense Insights MCP Server

通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。

官方
精选
本地
TypeScript
VeyraX

VeyraX

一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。

官方
精选
本地
graphlit-mcp-server

graphlit-mcp-server

模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。

官方
精选
TypeScript
Kagi MCP Server

Kagi MCP Server

一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。

官方
精选
Python
e2b-mcp-server

e2b-mcp-server

使用 MCP 通过 e2b 运行代码。

官方
精选
Neon MCP Server

Neon MCP Server

用于与 Neon 管理 API 和数据库交互的 MCP 服务器

官方
精选
Exa MCP Server

Exa MCP Server

模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。

官方
精选