Skip to main content
Many teams build internal AI agents — Jira bots, code-generation pipelines, automated PR reviewers — that commit code, open PRs, or comment under a dedicated service account. By default Weave treats those identities as human contributors. Custom AI agents let you register them so their activity rolls up into AI charts (AI output, AI output percentage, AI reviewer comments, etc.) instead of skewing your human metrics.

​
Who can use it?

  • Admin and Manager roles can create, edit, and delete custom AI agents
  • Custom AI agents apply organization-wide as soon as they’re saved

​
Setting up an agent

  1. Go to Settings → Custom AI agents
  2. Click Add custom AI agent
  3. Fill out the fields below and Save
After saving, you can edit the agent to upload a custom icon (JPG, PNG, or WebP). The icon shows up alongside built-in tools like Claude Code and Cursor wherever agents are listed.

​
Fields

​
Registrations

A registration ties the agent to a specific identity on a code platform. You need at least one. Each registration has:
  • Platform — GitHub, GitLab, Bitbucket, or Azure DevOps
  • Email — the exact git email the agent commits under (e.g. jira-ai@acme.com)
  • Email pattern — a wildcard for matching multiple addresses (e.g. *@bots.acme.com)
  • External ID — the platform-specific user ID (e.g. a GitHub numeric user ID)
You can provide any combination of email, email pattern, and external ID — Weave matches activity against whichever fields you set. Most teams start with only the email and add an external ID later if the agent ever commits under a different address.
Add a separate registration for each platform the agent operates on. A single agent can have one registration for GitHub and another for GitLab.

​
git-ai tool names

If you’re using git-ai for line-level AI attribution, list the tool identifiers your agent reports (for example jira-ai, internal-codegen). Names are case-insensitive. Lines attributed to any of these tools are credited to this agent.

​
How matching works

When Weave ingests a commit, PR, or review comment, it checks the author’s git email and platform user ID against every registration in your organization. The first matching agent wins, and the contribution is reclassified as AI activity attributed to that agent. This affects: Existing data is reclassified on the next analysis run — you don’t need to re-sync.

​
Editing or removing an agent

From the Custom AI agents list:
  • Click an agent to edit its name, description, category, registrations, or git-ai tool names
  • Click the delete icon to remove it — past activity reverts to being counted as the underlying human identity
Last modified on June 12, 2026