MCP server

The PRYSM Model Context Protocol server lets any MCP client — Claude Desktop, Cursor, Windsurf, and others — route prompts to the best-value model among 30+, preview costs, estimate savings, validate a BRAIN.md, and read usage. Your agent shouldn’t burn a frontier model on a one-line task; point it at PRYSM and it can pick the cheapest model that does the job well — no provider keys to juggle, one bill, one receipt per call.

Install

No global install needed — your MCP client launches it on demand with npx.

Claude Desktop
Cursor
Windsurf

Add to claude_desktop_config.json:

{
  "mcpServers": {
    "prysm": {
      "command": "npx",
      "args": ["-y", "@prysmai/mcp"],
      "env": {
        "PRYSM_API_KEY": "prysm_sk_your_key",
        "PRYSM_BASE_URL": "https://api.prysm1.com/v1"
      }
    }
  }
}

Add to ~/.cursor/mcp.json (or Settings → MCP → Add):

{
  "mcpServers": {
    "prysm": {
      "command": "npx",
      "args": ["-y", "@prysmai/mcp"],
      "env": { "PRYSM_API_KEY": "prysm_sk_your_key" }
    }
  }
}

Add to ~/.codeium/windsurf/mcp_config.json — same mcpServers shape:

{
  "mcpServers": {
    "prysm": {
      "command": "npx",
      "args": ["-y", "@prysmai/mcp"],
      "env": { "PRYSM_API_KEY": "prysm_sk_your_key" }
    }
  }
}

Restart your client after editing the config; “prysm” then appears as a tool provider.

Tools

Tool	What it does
`prysm_route`	Preview which model PRYSM would pick + estimated cost — no model call.
`prysm_chat`	Route a prompt to the best model and return the answer (+ model, cost, proof).
`prysm_orchestrate`	Cross several models (cascade / ensemble / debate) for one robust answer + agreement score and PrysmProof. Accepts a `compliance` spec to confine the run to approved providers.
`prysm_code`	Generate code with a self-correcting coder + critic loop — returns reviewed, repaired files + PrysmProof.
`prysm_agent`	Run a GOAL as a cost-bounded, autonomous multi-step agent (hard `max_cost_usd` ceiling enforced by AgentGuard). Returns the result + a verifiable PrysmProof v2 trajectory.
`prysm_compliance_preview`	Dry-run the Policy-as-Code compliance gate — no model call. Shows the data classes a prompt triggers, which models are allowed vs excluded under the org policy, the Compliance Cost Premium, and a sample attestation.
`prysm_estimate_savings`	Estimate how much you save by routing through PRYSM vs. always using a single reference model. Returns per-query breakdown + monthly projection at configurable query volume.
`prysm_verify_proof`	Verify a PrysmProof receipt by hash or request ID. Returns validity, model, provider, mode, and timestamp.
`prysm_models`	List the full model catalog with live pricing, tier, and context window.
`prysm_brain_validate`	Validate a `BRAIN.md` routing config against the catalog + schema.
`prysm_usage`	Usage stats for the configured key: requests, tokens, cost.

A great agent pattern: call prysm_route first to see the cost, then prysm_chat to execute — so the agent reasons about spend before committing to a model.

Configuration

Env var	Default	Description
`PRYSM_API_KEY`	—	Your `prysm_sk_*` key (required for `prysm_chat` / `prysm_usage`).
`PRYSM_BASE_URL`	`https://api.prysm1.com/v1`	API base URL (point at a self-hosted instance if needed).

The server communicates over stdio — the standard MCP transport. All diagnostics go to stderr, keeping the protocol channel clean.

What your agent gains

Cost-aware by default

The agent previews the price of a task and routes to the cheapest capable model.

One key, every model

27 models across 12 providers behind a single PRYSM key — no per-provider setup.

Governed by BRAIN.md

The same BRAIN.md rules and guardrails apply to agent calls.

Verifiable

Every prysm_chat returns a PrysmProof of what ran.

Get Started

Core Concepts

SDKs & Tools

Guides

Reference

Install

Tools

Configuration

What your agent gains

Cost-aware by default

One key, every model

Governed by BRAIN.md

Verifiable

​Install

​Tools

​Configuration

​What your agent gains

Cost-aware by default

One key, every model

Governed by BRAIN.md

Verifiable

Install

Tools

Configuration

What your agent gains