Skip to content

Integration

MCP Server

Connect your AI agents to TrustyGorilla via the Model Context Protocol. Let them read flag state, evaluate targeting rules, and safely toggle features — all through natural language.

  • Works with Claude, Cursor, Copilot, and any MCP-compatible agent
  • Read and write flags from within your AI workflow
  • Scoped tokens — agents only touch what you allow
  • Every action logged in your existing audit trail
Get startedMCP docs
AI agent session connected
Enable the new checkout flow for 10% of users in production

Tool call

enable_flag({
  key: "new-checkout",
  env: "production",
  rollout_pct: 10
})

Result

{
  "ok": true,
  "flag": "new-checkout",
  "env": "production",
  "rollout_pct": 10,
  "propagated_in": "14ms"
}
AI
Done. new-checkout is now live in production at 10% rollout. Propagated to all edge nodes in 14ms.
8
MCP tools exposed
SSE
Streaming transport
100%
Audit logged

Available tools

What agents can do

The TrustyGorilla MCP server exposes a set of typed tools your agents can call. Each tool validates inputs, enforces permission scopes, and writes to the audit log.

Full tool reference
MCP tools
  • get_flagEvaluate a flag for a given user context
  • list_flagsList all flags in a project with current state
  • enable_flagEnable a flag in a specific environment
  • disable_flagDisable a flag in a specific environment
  • get_configFetch a remote config value by key
  • set_configUpdate a remote config value
  • get_experimentFetch experiment status and results
  • list_audit_logQuery recent changes for a flag or project

Quick setup

Connect in two steps

1Add to your MCP config
{
  "mcpServers": {
    "trustygorilla": {
      "command": "npx",
      "args": ["-y", "@trustygorilla/mcp"],
      "env": {
        "TG_API_KEY": "your-api-key"
      }
    }
  }
}
2Start using it

Restart your MCP client. TrustyGorilla tools will appear in the tool list automatically.

Ask your agent anything about your flags:

"What flags are enabled in production?"
"Roll back the dark-mode flag to 0%"

Capabilities

Built for AI workflows

AI-native flag control

Expose your flags as MCP tools. Any compatible AI agent can evaluate, enable, or disable features through natural language.

Real-time context

Agents always see live flag state — no stale snapshots. Decisions are grounded in what's actually running in production.

Scoped permissions

Issue MCP tokens with read-only or write access per environment. Agents can never touch flags they're not authorised for.

Audit every action

Every flag change an agent makes is logged in the same audit trail as human changes — with the agent's identity attached.

Give your agents real-time flag control

Connect TrustyGorilla to your AI workflow in two minutes.

Start freeMCP reference
  • No credit card required
  • Free forever tier