2026-09-30

Which model does Sume Agent Completions run? Only sume-agent

The model field on POST /v1/agent/completions accepts only sume-agent. Omit it for the same agent; any other value returns 400 invalid_request.

4 min readSume

All posts

The model field on POST /v1/agent/completions accepts only sume-agent, and omitting it gives you the same agent. Any other value returns 400 invalid_request, so you cannot pick a different model per call.

Everything here is from Agent Completions, read 2026-09-30.

What does the receipt show?

The 202 receipt reports model: "sume-agent" and object: "agent.run", with a status of queued at first. The docs describe the runtime as the same one behind the Agents chat UI: sandbox, tools, MCP bridge and media generation.

What does the request accept?

Request fields from the Agent Completions docs, read 2026-09-30: https://docs.sume.com/agents/completions
Field	Required	Notes
`instruction` or `messages`	One of	Never both
`generation_spend_cap_usd`	Yes	No default; omitting it fails with `400 invalid_request`
`model`	No	Only `sume-agent`
`input`	No	Caller data, treated as data
`output_schema`	No	Bind `output` to your schema
`communication.webhook_url`	No	Public HTTPS URL for the terminal event

What varies per call instead of the model?

Not on this endpoint. The task itself varies per call: you send an instruction or messages and a spend cap, and the Agent runs with its tools and media generation.

How do I set a minimal request?

Send the task and a spend cap. The cap is the per-run ceiling for generation spend, and the docs suggest sizing it against the metered rates on the API pricing page.

curl -sS -X POST "https://api.sume.com/v1/agent/completions" \
  -H "Authorization: Bearer $SUME_API_KEY" \
  -H "Content-Type: application/json" \
  -H "Idempotency-Key: model-check-001" \
  -d '{
        "model": "sume-agent",
        "instruction": "Write a two-line product teaser.",
        "generation_spend_cap_usd": 1
      }'

Which scopes and keys can call it?

The key needs agent_completions:write to create a completion or cancel a run, and agent_completions:read to read and list runs. An older key fails every request with 403 insufficient_scope, and you create a new key and rotate to it. Service-account keys cannot create completions. Unknown or foreign run ids return 404 agent_run_not_found.

Sources

Agent Completions