FAQ
How it works
No. Start from a template for payments, CRM, finance, cloud or coding agents, test a sample agent call in the editor, and run new rules in shadow mode before they block anything.
Keep your framework and models. Runledge evaluates proposed tool calls, coordinates human approvals and records outcomes through REST or OAuth MCP. Your executor honors those decisions and calls the provider.
It plans the undo; your systems carry it out. Agents can declare an undo step with each action. When you start a rollback, Runledge returns those steps in reverse order and tracks each one until it is done. It does not call providers itself or guarantee exactly-once effects.
Create an account, confirm your email and create a workspace. Configure a policy, then use a workspace API key for REST or authorize a remote MCP client through OAuth. Check before calling your provider and record the confirmed outcome afterward. No enabled rules means proposals are allowed, so configure your policies first.
Security and deployment
Runledge stores submitted action metadata, approval proposals, policies and audit records. It does not need your model prompts or provider credentials for pre-flight checks. Send only the fields needed for your policies; zero-retention processing is not currently available.
The source includes a container build and Kubernetes manifests with an MCP transport sidecar. Customer-hosted deployment requires a compatible, independently operated backend and auth service, runtime secrets and release validation. It is not currently an air-gapped appliance or a published container image.
Pricing and Runledge AI
Your existing rules check proposed actions. Runledge AI reviews recorded activity to find policy gaps and volume spikes, drafts candidate rules and links findings to the actions behind them. You review and apply any rule changes.
Your agent’s model decides what to attempt. Runledge AI analyzes the activity recorded in your workspace to suggest better policies. Your configured rules still determine whether a proposal is allowed, blocked or held for approval.
Yes. Team includes Runledge AI, and Growth and Enterprise inherit it. Growth also adds multi-person approval and cost-center reporting.
Run a review in the Rules page or request recommendations through REST or MCP. Each review examines up to 5,000 actions from the last 30 days. Investigations show recorded activity, not a verified root-cause diagnosis.
Yes. Any rule can run in shadow on any plan: it records what it would have blocked or held without stopping the action. Enterprise teams can ask us for a guided rollout across their agents.
Developer has a hard cap of 1,000 actions per month. Team includes 25,000 and Growth includes 250,000; extra actions on those plans cost $3.00 per 1,000 and appear on your next invoice. Your agents never stop at the limit.
Around 100,000 actions a month, Team plus extra actions costs about as much as Growth, which includes 250,000 actions, multi-person approval and cost-center reporting.
Yes. Enterprise contracts include a committed action volume at a lower rate, plus in-VPC deployment, SCIM and a custom SLA. Contact sales@runledge.com.