The Supreme Mind API
Run the AI Agents from your own tools: REST API and MCP server. This is the Expert Layer of the Legal AI Stack in practice: the same Expert Agents as the app, called from your own product or agent. Pick a class of expert, point it at a matter, and get the brief back as structured data, with the verification status of every authority attached.
The API is in early access. Keys are issued by invitation while the first partners are onboarded; to ask for one, write to us or book a demo.
Who it is for
- AI-native law and litigation firms running their own matter workflow and agents, who want each side’s expert analysis without leaving it.
- Legaltech platforms that want expert agents as a feature for the litigation teams already working in their product.
- AI labs and agent builders connecting legal agents to a tool whose quotations are checked against the public record.
What you can call today
| Endpoint | What it does |
|---|---|
GET /v1/expert-classes | The classes of expert available to your key, with their methods, typical questions and common attacks. |
POST /v1/matters | Create a matter from structured facts or plain text. GET and DELETE by id. |
POST /v1/briefs | Start a brief. It runs in the background and answers at once with an id; a brief takes several minutes. |
GET /v1/briefs/{id} | Status, then the brief: likely opinion, cross-examination weaknesses, the Rule 702 attack surface and the settlement range, with each citation's verification status. |
GET /v1/usage, GET /v1/audit-log | Usage per key, and your account's audit events. |
- Webhooks.
brief.completedandbrief.failed, signed to the Standard Webhooks scheme, so you can stop polling. - Safe retries. Send an
Idempotency-Keyon any POST and a repeat returns the first answer instead of a second brief. - Test keys.
sm_test_keys return a fixed sample brief, never call a model and never charge, so you can build the integration first. - Reference. The full contract is an OpenAPI 3.1 document at api.suprememind.ai/v1/openapi.yaml.
For agents: the MCP server
Claude and other MCP clients can call the same AI Agents as tools. Add https://mcp.suprememind.ai/mcp as a custom connector and sign in with your Supreme Mind account; approving the connection needs API access to be switched on for your firm. Agents running on your own servers can use an API key instead.
- Read-only tools:
list_expert_classes,get_expert_class,get_brief,get_job,get_usage. - Matters:
create_matter, free. - Metered tools:
create_brief, charged when the brief is delivered, andstart_cross_examwithcross_exam_turn, a cross-examination practice session charged at its start. - The expert’s data sources: every brief runs on them: an event study and SEC EDGAR for securities experts, a cost pass-through study for antitrust experts, and the judicial record on the method for all. Pass
groundings(ticker, event dates, class or conspiracy period) when the facts name them; otherwise the sources the matter supports run. The brief reports which ran. - Each tool description carries its cost and use limits, so the agent sees them before it acts. Sign-in is OAuth 2.1 with PKCE, scoped to
read,briefsandcross_exam. - Full specification: authentication, scopes, every tool, argument and result, in the MCP server specification (PDF).
Coming next
- Document upload and fact extraction through the API, and the brief as a Word document.
- Expert report analysis (
analyze_report) through MCP. It is available today in the app and the REST API.
What every brief carries
Every citation comes with the result of its check against the public record: verified, found by name, or unconfirmed. Every brief also carries a provenance block stating what was checked (quotations and citations) and what was not (figures, calculations and the analysis). Integrations show both to the lawyer reading the output; that is a condition of access, because a label that is stripped on the way through protects nobody. How the checks work is set out in how it works.
How the output may be used
- Attorney work product and consulting analysis only. It is not evidence and it is not a testifying expert’s opinion.
- A licensed lawyer reviews it before anything is relied on, and before any of it reaches a testifying expert.
- Every expert agent is a composite of a class. No output names, or presents itself as, a real expert.
- Platforms pass these terms on to their own users.
Data
The API runs on the same infrastructure as the app and keeps the same commitments: data stays in the United States, prompts and completions are not logged or stored, nothing you send is used to train models, and deleted matters are purged. Keys are stored only as hashes, scoped to what they may do, and can be limited to your own addresses. The detail is on the security page.
Pricing
On a firm plan, API and MCP access is included, with the same unlimited runs as in the app. Outside a plan, a brief is charged on your firm’s agreed terms when it is delivered. Test keys are free. Volume and platform terms are agreed with each partner; see the pricing page.
Early access. Features marked as coming next are in development and are not yet available. Supreme Mind produces internal preparation for counsel and is not legal advice.