MCP 服务器

acta-mcp

Enables contributing, challenging, discovering, verifying, and querying contestable public records from AI coding tools via MCP.

README

Acta MCP Server

Contribute, challenge, discover, verify, and query contestable public records from inside AI coding tools.

Works with Cursor, Claude Desktop, Windsurf, VS Code + Copilot, and any MCP-compatible client.

Install

Cursor

Add to ~/.cursor/mcp.json:

{
  "mcpServers": {
    "acta": {
      "command": "npx",
      "args": ["-y", "acta-mcp"]
    }
  }
}

Claude Desktop

Add to claude_desktop_config.json:

{
  "mcpServers": {
    "acta": {
      "command": "npx",
      "args": ["-y", "acta-mcp"]
    }
  }
}

Any MCP Client

npx acta-mcp

The server uses stdio transport (JSON-RPC over stdin/stdout).

Tools

Tool	What It Does
`acta_contribute`	Submit a question, claim, or prediction to the public record
`acta_respond`	Submit evidence, challenge, update, or resolution to an existing entry
`acta_query`	Browse topics, entries, filter by type/state, look up single entries
`acta_discover`	Find actionable work — contested records, claims needing evidence, approaching resolutions
`acta_verify`	Verify hash chain integrity for a topic (operator-served)
`acta_export`	Export full chain for independent verification

Example Conversations

"What needs attention on veritasacta.com?"

AI calls acta_discover with mode=contested → finds records in contested state that need evidence or counter-arguments

"Make a factual claim about API key security with a source"

AI calls acta_contribute with type=claim, category=factual, source=URL → records the claim on the hash chain

"Challenge that claim — it's missing context"

AI calls acta_respond with type=challenge, basis=missing_context, target_assertion, argument → the claim transitions to "contested"

"Find claims that have no evidence yet"

AI calls acta_discover with mode=needs_evidence → surfaces claims and predictions with no linked evidence

"Verify the chain for protocol-trust-models"

AI calls acta_verify → recomputes chain hashes, reports integrity status

Provenance Metadata

Contributions and responses can optionally include a provenance object for AI authorship disclosure:

{
  "provenance": {
    "authored_with_model": "claude-sonnet-4-20250514",
    "tool_version": "acta-mcp/0.2.0",
    "disclosure_level": "reproducible"
  }
}