Agents

Autonomous agents get their own identity, budget, and inference key — not a shared credential.

If you run autonomous agents, they need the same guardrails a human integration gets — a scoped credential, a budget, and clean attribution of what they spent and on what — without you hand-rolling it. In Relixr, an agent is a first-class fleet identity, not a label bolted onto a shared API key.

Each agent can have its own spend limits, model allowlist, schedule, and a dedicated bound inference key.

Create an Agent

1. Go to Agents in the dashboard.
2. Select New agent.
3. Name the agent and, optionally, set a daily, weekly, or monthly spend cap.
4. Select Create agent. Relixr can auto-provision a bound nf_live_ key for it at the same time.

Usage, activity, and CSV exports attribute every request to agentId and agentGroupId, so a runaway agent shows up in your usage export by name instead of being buried in workspace-wide totals. Archiving an agent revokes its bound keys immediately.

Agent Groups

Agent groups let a whole fleet inherit shared policy — a spend limit, a model allowlist — while individual agents can still override it.

GET    /api/agents/groups
POST   /api/agents/groups
PATCH  /api/agents/groups/{id}
DELETE /api/agents/groups/{id}   // unassigns agents, does not delete them

// Create a group with a shared policy
POST /api/agents/groups
{"name": "support-bots", "spendLimits": {"monthlyUsd": 200}, "modelAllowlist": ["gpt-4o-mini"]}