Skip to content

A Savvagent service

Savvagent Flags

Feature flags that your agents manage

Agents create, target, roll out and roll back flags through MCP tools, while your running apps evaluate them through SDKs and REST, with no LLM in the hot path.

Savvagent Flags works inside your AI assistant; it has no web console of its own. Usage and billing are in your Savvagent console.

What it does

  • Create flags, set targeting rules and raise rollouts in steps, all through MCP tool calls.
  • Assess the risk of a rollout and correlate a flag with error spikes.
  • Roll back to any earlier version of a flag in one call.
  • Apps evaluate flags through SDKs and REST, so a production request never waits on an LLM.
  • Automatic rollback on Team plans and above.

Connect your agent

Add this address to your AI assistant or coding agent as a custom connector, and sign in with your Savvagent account when it asks.

Connector address
https://otto-flags.savvagent.com/mcp

Claude (claude.ai or the desktop app)

Open Settings, then Connectors, and choose Add custom connector. Paste the connector address, select Connect, and sign in with your Savvagent account.

ChatGPT

Turn on developer mode, then open Settings, then Connectors, and choose Create. Paste the connector address and sign in with your Savvagent account.

Claude Code

Command
claude mcp add --transport http otto-flags https://otto-flags.savvagent.com/mcp

Then run /mcp in Claude Code and sign in with your Savvagent account.

See pricing: pay as you go, per service

Other Savvagent services