AI agents your organisation can answer for.

Governed by design

The point of an agent is that it does real work. That only works if it can be held to the same rules as a colleague.

• Agents are users

  Each agent runs as its own user in Copyl, with its own permissions. It sees and changes what that user may see and change — nothing more.

• Tools you switch on

  Choose per agent which actions it may use, from the catalogue of Copyl's integration platform. An agent without a tool cannot take that action.

• A person approves

  Consequential actions become approval requests. When someone approves, the action runs, and the request shows whether it succeeded.

• Full action trail

  What each agent did, when, and on whose behalf is recorded and visible in the activity view and the audit tab.

• Handover to a person

  When an agent should not continue on its own, it hands over: the conversation says so and a task is created for a person.

Run it, measure it, improve it

An agent is not finished when it is configured. Agent Hub keeps it running and shows whether it is getting better.

• Triggers and schedules

  Run an agent on a schedule, on an event, or by hand, with a history of every time a trigger fired.

• Choose and swap models

  Pick the model per agent. Swap models across agents with a replay comparison first, then promote or roll back.

• Test cases and review

  Keep test cases per agent and review outcomes. Prompt improvements are proposed from real outcomes, and learnings are kept with the agent.

• Versions

  Every saved configuration is a version you can compare and restore.

• Usage and cost

  Execution logs and usage per agent, with the cost of each run in Copyl Work Units and a monthly total for the organisation.

• An assistant in every app

  Each Copyl app has its own assistant that knows the app it sits in, built and governed in Agent Hub like any other agent.

Built elsewhere? Govern it here.

Teams build agents in many places. Agent Hub does not ask them to stop — it brings those agents under one set of rules.

• Import Claude, OpenAI and MCP agents

  Bring an agent definition in, run it in a sandbox, approve its scopes, sign off and deploy — and pause or revoke it at any time. How import works

• See what n8n workflows do

  Connect n8n and see which workflows use AI, their runs and failures, and model spend per workflow. Copyl and n8n

• Alongside Microsoft

  Running Copilot Studio or Agent 365? Agent Hub governs agents regardless of which stack built them. Copyl and Microsoft

• Your own compute

  Run agent work on self-hosted Agent Workers with local models, monitored from Agent Hub.

Plans

Flat price per organisation, not per user. Billed yearly by default, monthly if you prefer. Prices excl. VAT.