Sidclaw Governance Mcp
Wrap risky MCP tools behind policy checks, human approval, and audit logs before agents act in production.
Overview
SidClaw Governance MCP is a MCP server for the Ship phase that proxies upstream MCP tools through policy evaluation, optional human approval, and audit trails.
What is this MCP server?
- Stdio proxy in front of any upstream MCP server via SIDCLAW_UPSTREAM_CMD and SIDCLAW_UPSTREAM_ARGS
- Policy evaluation per agent using SIDCLAW_AGENT_ID and SidClaw API
- Human approval modes: error (fail fast) or block (wait for approval)
- Audit logging tied to SIDCLAW_API_KEY for traceable agent actions
- npm package @sidclaw/sdk at server version 0.1.7
- Server schema version 0.1.7 on npm package @sidclaw/sdk
- 5 documented environment variables including required API key, agent ID, and upstream spawn config
- 2 approval modes: error (default) and block
Community signal: 11 GitHub stars.
What problem does it solve?
Agents can call powerful MCP tools with no consistent policy, approval, or audit record when you chain multiple servers in production.
Who is it for?
Indie builders running multiple MCP integrations who need a single governance layer before granting agents production credentials.
Skip if: Local-only experiments with no sensitive data, or teams that cannot use a SidClaw API key and hosted policy service.
What do I get? / Deliverables
After install, tool calls flow through SidClaw so allowed actions proceed, risky ones can pause for approval, and decisions leave an auditable trail per agent.
- Proxied MCP tool surface with policy checks on each call
- Configurable approval behavior and agent-scoped audit records
Recommended MCP Servers
Journey fit
How it compares
Governance proxy MCP layer, not a domain integration or agent skill.
Common Questions / FAQ
Who is SidClaw Governance MCP for?
Solo and small-team builders who already use stdio MCP servers and want policy, approval, and logging without rewriting each tool.
When should I use SidClaw Governance MCP?
Use it during Ship when you connect agents to real systems and need to block or approve high-risk tool calls centrally.
How do I add SidClaw Governance MCP to my agent?
Install @sidclaw/sdk, set SIDCLAW_API_KEY, SIDCLAW_AGENT_ID, SIDCLAW_UPSTREAM_CMD, and SIDCLAW_UPSTREAM_ARGS, point your agent at this server as the MCP entry, and choose SIDCLAW_APPROVAL_MODE if you need blocking approval.