Skip to main content

Caller identity

By default, every call to your data source looks like it came from incident.io. Turn on Share caller identity and each call tells your server who actually made it, so you can:
  • Apply access controls per caller.
  • Rate-limit each caller separately.
  • Restrict what automated callers, such as investigations, can do.
  • Log who made each call.
It works with MCP and HTTP connectors and some telemetry sources.

Turn it on

Open the data source’s settings and turn on Share caller identity. You need the Manage Nexus sources permission. We record the change in your audit log. From then on, every call to that source carries the caller, whether it comes from Investigations, Chat, a workflow or the incident.io MCP.

What your server receives

Each request gets two extra headers:
  • X-Incident-Actor says who the caller is. It matches the actor on the call’s audit log entry.
  • X-Incident-Actor-Assertion is a signed JWT that proves it. Verify it before you trust the header.
The caller is one of:

What’s in the assertion

Here’s a decoded assertion for a user who asked Chat a question during an incident, where someone had to approve the call first:
And one for an investigation, which runs with no user behind it:
ctx.surface is always set, for example chat, investigation or mcp_client. The other keys only appear when they apply:

Verify the assertion

Fetch our public keys from https://app.incident.io/.well-known/jwks.json, pick the one matching the assertion’s kid, and check the signature, issuer, audience and expiry. Then check sub matches X-Incident-Actor. For example, in Node.js with jose:
The keys endpoint needs no authentication, and you can cache it for 5 minutes. Don’t hard-code a key or kid: we rotate them.