The key is configured by the owner. The agent cannot create an account or issue itself credentials.
WHAT THE AGENT CAN DO
Research first. Review before merge.
ask_accord
Research questions and engineering plans
Four systems investigate independently, then Accord resolves the evidence into one answer or implementation plan. Supply relevant files for repository-grounded planning.
RETURNS answer · sources · usage
Request and response example
REQUEST
ask_accord({
"query": "Research this architecture and produce a risk-ranked implementation plan.",
"context": [{ "path": "docs/architecture.md", "content": "…" }],
"strategy": "accord"
})
Specialized reviewers examine correctness, security, tests, and operational risk. A final verifier rejects unsupported claims and merges duplicate findings. Your coding agent supplies the diff and relevant context; Accord does not clone the repository.
RETURNS verdict · verified findings · file and line
Request and response example
REQUEST
review_code({
"task": "Review the current change before merge",
"diff": "<git diff collected by the coding agent>",
"focus": ["correctness", "security", "tests"]
})
Subscribe to Accord, create an API key, and use included usage across every model. Card overages remain optional and capped by the account owner.
AVAILABLE
Accord API key
The owner subscribes and creates an accord_… key. Included usage is used first; optional overages are charged to the saved card up to the owner's monthly cap.
Use the answer endpoint for research and planning, and the review endpoint for an evidence-based verdict on a supplied change. Answer streaming is optional; the low-level review endpoint always returns NDJSON.