System prompt

Give any agent the Chutes operating rules.

For agents without a first-class integration, use the toolkit system prompt as the source of truth for URLs, auth, live model discovery, routing caveats, and TEE boundaries.
Open source guideConnect quickstarts
System prompt sourceConnect draftOpenAI-compatible guide
Status
Generic prompt lane
Base URL
https://llm.chutes.ai/v1
Live model example
Qwen/Qwen3-32B-TEE
Catalog checked
Jun 25, 2026; context 40K

Full setup walkthrough

Everything you need is on this page — no need to leave for the toolkit. Each step is copy-paste runnable and uses a live model id from the catalog.

  1. 01

    Paste the operating rules into the system prompt

    Start from the toolkit system prompt rather than hand-written endpoint notes, so URLs, auth, routing, and TEE boundaries stay correct:
    system prompt core
    text
    You have access to Chutes.ai inference.

Use https://llm.chutes.ai/v1 as the OpenAI-compatible base URL.
Use Authorization: Bearer cpk_... for inference.
Read GET /v1/models before choosing model IDs, prices, context, features, or TEE claims.
Use confidential_compute, not a model-name suffix, for TEE selection.
Do not claim cryptographic attestation unless DCAP verification ran and passed.
  2. 02

    Keep model facts live

    Instruct the agent to read GET https://llm.chutes.ai/v1/models before stating any model id, price, context window, feature flag, or TEE claim.
  3. 03

    Preserve the boundaries

    Bearer auth everywhere, the routing-alias setup caveat, every [BETA] label, and the TEE honesty rule (confidential_compute boolean; no cryptographic-attestation claim without DCAP).

Rules worth preserving

Prompt rule
Use Authorization: Bearer cpk_... everywhere.
Why
X-API-Key is silently ignored on inference and can fall into anonymous limits.
Prompt rule
Read /v1/models before stating IDs, prices, context, features, or TEE flags.
Why
The live API is the source of truth; examples can go stale.

Troubleshooting

Symptom
The model ID fails
Likely cause
The catalog changed or the client cached model metadata.
Fix
Refresh /v1/models and update the config.
Symptom
Auth falls into anonymous limits
Likely cause
The key is missing or not sent as Bearer auth.
Fix
Export CHUTES_API_KEY=cpk_... and send Authorization: Bearer.
Symptom
Routing aliases fail
Likely cause
Saved default aliases need a dashboard Model Routing pool.
Fix
Use a concrete model ID or an inline comma-separated pool first.