Developers
Governance without a rewrite.
IntSen governs your agent's model calls at the gateway, so the integration lives in configuration, not in your agent's logic.
What you change
- The model client's base URL → the IntSen AI Gateway
- The API key → an OAuth client-credentials token for your agent
- A few headers: registration, agent, a task id per user message
- Two YAML files: a Registration and an Agent
What stays the same
- Your agent framework and its version
- Your prompts, tools and agent loop
- Where your agent runs—IntSen never hosts it
- Request and response formats of the provider API
Reference
The headers
Two are set once on the client. Two or three are set per conversation or turn.
| Header | Set | Value |
|---|---|---|
| x-intsen-registration-id | Once | Your registration name |
| x-intsen-agent-id | Once (optional) | namespace/agent-name; IntSen can infer it from the login |
| x-intsen-task-id | Every user message | A new ULID or UUID per turn |
| x-intsen-conversation-id | Per conversation (recommended) | Your chat, thread or session id |
| x-intsen-user-id | Per conversation (optional) | The end user; stored only as a hash |
IntSen headers are removed before the request reaches the model provider.
Tooling
intsenctl and the intsen package
intsenctl applies resources from your laptop or pipeline, with dry runs. The optional intsen Python package helps run golden evaluation tests from CI. Neither is needed at runtime by your agent.
terminal
pip install intsenintsenctl apply -f registration.yaml -f agent.yaml --dry-runGuides
Developer documentation
Talk to an engineer
Book a technical walkthrough and we'll connect one of your agents with your team.