TaskGate Docs
taskgate.ai Request access
Overview

What TaskGate is

TaskGate sits between your AI agents and the systems they act on. Every consequential action an agent proposes is decided by your policy before it runs, and every decision is recorded.

The gate

When an agent wants to call a tool — send a payment, post a message, update a record — TaskGate turns the request into a proposed action and decides it against the policies in force. There are three outcomes:

AllowThe action runs.
Require approvalThe action waits until the right people approve it.
RefuseThe action never runs. The reason goes back to the caller.

The decision is made again, against fresh data, at the moment the action executes. An approval that no longer satisfies the policy at that point does not let the action through.

Default

Until a policy set is in force, the gate refuses by default.

What you work with

PoliciesThe rules every proposed action is decided against. ApprovalsThe inbox where people decide on held actions. Audit trailA per-organisation hash chain with signed checkpoints. Agents & identityRegistered agents, scoped credentials, delegated authority. Budgets & costHard and soft spend limits, and a cost ledger. Integrations100+ connectors, all brokered through the gate. Governed chatChat with your own Claude under the same policy and audit. Evaluation & replayTest agents before release and replay any run.

Who uses TaskGate

Signing in

The console is at the sign-in screen Console · sign in. Enter your organisation's short name in the Organisation field and choose Continue to sign-in. You are sent to your organisation's identity provider to sign in, and the console remembers the organisation on that device.

New to TaskGate? Start with Getting started.