Agent access

Agent access lets you connect MCP-compatible repair agents to Errorgap with scoped tokens. Tokens are safer than sharing a user browser session because they can be limited by team, project, capability, and expiration.

What agent access does

A token allows an agent to read selected evidence through the MCP endpoint. Depending on the capabilities you grant, the agent can inspect errors, source context, logs, and proposed fixes.

Create a token

1. Open Organization, then Agent access.
2. Copy the MCP endpoint if you need to configure a client manually.
3. Select New token.
4. Name the token after the agent or workflow that will use it.
5. Choose scope, capabilities, and expiration, then create the token.

Copy the token when it is created. Errorgap does not display the full token again later.

Choose scope

Scope controls what the token can reach. Start narrow and expand only when the agent needs broader context.

• Team scope limits the token to projects visible to that team.
• Project scope narrows access further when one agent should work on one project.
• Expiration limits how long the token remains usable.

Choose capabilities

• Read errors allows the agent to inspect groups and occurrences.
• Read source allows source context for stack frames and proposed fixes.
• Read logs allows log evidence near errors and incidents.
• Write fixes should be granted only when the agent is expected to open or modify fixes.

Connect an agent

Use the token detail page to copy the client configuration for your agent. Confirm that the agent can reach the MCP endpoint and that project privacy settings allow the evidence categories it needs.

Rotate and revoke

Revoke tokens when an agent is retired, a token is exposed, a teammate leaves, or the scope is no longer appropriate. Create a replacement token instead of reusing a broad or expired one.