The Routescore API & MCP server.
Pre-sign swap checks are free for agents — 100/day on any account via the REST API + MCP, no card. Pro raises it to 1,000/day, Power to 10,000/day. Power also unlocks the metered compute endpoints (modeled premium estimates + what-if simulation). Drop Routescore tools straight into Claude, Cursor, and your agents. Read-only public snapshots and account exports live below.
Call check_swap before signing an onchain swap.
One read-only pre-sign evidence read: a modeled route-quality grade and modeled slippage estimate, modeled MEV/execution exposure where observable, and a token registry recognition read, composed into a clear / caution / unsupported verdict with reasons and caveats. It does not execute, route funds, sign, or promise an outcome — it is the evidence element your planner, executor, and wallet compose around. The modeled figures are backed by a published methodology and calibration state, not adjectives.
Drop @routescore/mcp into Claude Desktop or Claude Code — two lines of config.
Add @routescore/mcp to ~/.codex/config.toml — Codex loads MCP servers like Claude.
A function-calling tool definition or a Custom GPT Action from the OpenAPI spec.
A tool-use definition on xAI’s OpenAI-compatible API.
A function-calling declaration for the Google GenAI SDK.
Building the agent itself? See add a pre-sign evidence check and the agent skills for the relay discipline. Read-only decision support; not investment advice, and not an execution service.
Call Routescore from your own stack
- Generate a key in Account → Developer → API & MCP — free on any tier; the key inherits your plan's limits.
- Send it as Authorization: Bearer rs_live_… to any endpoint below, or
- Drop the MCP config into your client to expose the same tools to your agent.
Know which base URL and capability state you are calling
| Environment | Base URL | State | Boundary |
|---|---|---|---|
| Production | https://www.routescore.io/api/public/v1 | partial | Existing keyed pre-sign and read-only surfaces use the published OpenAPI contract. Production remains on MCP 0.3.5 with eight tools and filters stored-execution REST/OpenAPI paths. Stored-record execution review, public-transaction reconstruction, and compiler-backed tenant reports are source-tree/DEV-preview capabilities only until the production schema is migrated, verified, and the server-only release gate is explicitly enabled. |
| Local development | http://localhost:3000/api/public/v1 | partial | Behavior depends on local fixtures, database, and source configuration. |
| Public testnet | Not published | unavailable | No separate public testnet API is claimed. |
Pre-sign swap check
Free agent tier: 100 calls/day (Pro 1,000, Power 10,000). Agent/wallet preflight: chain support, route quality, token registry recognition, caveats, and clear/caution/unsupported verdict. Not execution or advice.
MEV premium estimate
Modeled premium estimate + expected/CVaR loss for sandwich exposure on a swap. Not a live cover offer.
Bridge premium estimate
Modeled premium estimate for cross-chain bridge execution-failure exposure. Not a live cover offer.
LRT slashing premium estimate
Modeled slashing-exposure premium estimate for an LRT position given AVS exposure. Not a live cover offer.
What-if scenario
Monte-Carlo modeled premium estimate vs above-threshold loss over a horizon.
Preflight evidence records
Newest-first index of the hash-verifiable evidence records your keyed check/swap calls persisted. Owner-scoped: a key only sees its own account’s records.
Preflight evidence record
One persisted record by the record_id check/swap returned: the original response verbatim plus a canonical-JSON SHA-256 integrity hash, re-derivable offline.
Detector manifest
Latest public detector run: version hash + eligible universe.
Whoami
Verify your key works and see its plan tier.
curl -X POST https://www.routescore.io/api/public/v1/quote/mev \
-H "Authorization: Bearer rs_live_..." \
-H "Content-Type: application/json" \
-d '{"notional_usd": 10000, "asset_pair": "USDC/ETH"}'curl -X POST https://www.routescore.io/api/public/v1/check/swap \
-H "Authorization: Bearer rs_live_..." \
-H "Content-Type: application/json" \
-d '{
"notional_usd": 10000,
"chain_id": 4663,
"route": "uniswap-v3-rho",
"token_in": "USDG",
"token_out": "0xaF3D76f1834A1d425780943C99Ea8A608f8a93f9",
"slippage_allowance_bps": 75
}'For RHC tokenized assets, pass the Routescore registry contract address. recognized means registry match only: not safe, sellable, redeemable, rights-bearing, or jurisdiction-eligible.
{
"mcpServers": {
"routescore": {
"command": "npx",
"args": ["-y", "@routescore/mcp@latest"],
"env": { "ROUTESCORE_API_KEY": "rs_live_..." }
}
}
}Published @routescore/mcp@latest is 0.3.5 with 8 tools, including the check_swap pre-sign check and get_preflight_record for retrieving persisted evidence records. The live hosted /api/mcp transport matches that 0.3.5, 8-tool surface. The source tree and repository manifest prepare 0.4.0 with 11 tools for DEV previews: owner-scoped record_execution and review_execution require an authenticated Pro key, while review_public_transaction is declared for contract discovery but is not enabled in production. Promoting the hosted candidate or publishing the 0.4.0 stdio npm package remains founder-gated. Rate limits apply per key (X-RateLimit-* headers).
{
"score_state": "partial",
"source_freshness": {
"state": "partial",
"checked_at": "2026-06-21T00:00:00.000Z",
"sources": [
{ "name": "routescore_backend", "freshness_state": "fresh" },
{ "name": "bridge_risk_labels", "freshness_state": "unknown" }
]
},
"methodology_version": "routescore.public_api.v1",
"confidence_band": { "low": null, "high": null, "unit": "bps" },
"caveats": [
"Modeled, point-in-time decision support. Not an execution guarantee.",
"Unsupported or stale inputs widen uncertainty instead of hiding risk."
],
"commercial_disclosure": {
"paid_placement": false,
"score_influenced_by_partner": false
}
}Premium estimates and scenarios are modeled, point-in-time, user-controlled decision support — not investment advice, and not a live cover, insurance, or premium-acceptance offer. The export surfaces below are read-only. Routescore does not custody assets, execute transactions, underwrite risk, or recommend financial actions. API and MCP responses include score state, source freshness, methodology version, caveats, confidence, and commercial disclosure so downstream tools can keep the trust contract visible. Known limitations are published at /limitations; the full DeFi risk disclosure is at /risk-disclosure.
One composed element of your agent stack
Routescore is the evidence/preflight element of an agentic onchain-finance stack — one composable element alongside your planner, executor, and wallet tooling. Every surface below is read-only decision support; nothing executes, signs, or moves funds.
Keyed REST API
The endpoints above — check/swap pre-sign verdicts with trust envelope, quotes, scenarios, and evidence records — from any backend. Guide · OpenAPI
@routescore/mcp
The published @latest is 0.3.5 with 8 tools for Claude, Cursor, and agent frameworks, including check_swap and get_preflight_record. The live hosted /api/mcp transport matches that 0.3.5, 8-tool surface. The source tree and repository manifest prepare 0.4.0 with 11 tools for DEV previews, adding owner-scoped Pro record_execution, review_execution, and production-disabled review_public_transaction (full tool table). Promoting the hosted candidate or publishing the 0.4.0 stdio npm package remains founder-gated. An evaluated unsupported verdict arrives as a structured result (gap-state fields and caveats intact), not as an error.
Copy-paste examples
MCP client configs (Claude Desktop, Cursor, Codex), REST examples in curl, TypeScript, and Python, and an offline verifier for the evidence record’s canonical-JSON SHA-256 hash — github.com/routescore/agent-examples.
Agent skills (routescore-ai)
Two SKILL.md files — check-swap-preflight and verdict-and-caveat-relay — served raw from this site; no registry needed. A relay linter ships in the MCP package. Install guide
Records + offline verification
Keyed check/swap calls attempt to persist an owner-scoped evidence record and return its record_id — null, with an explicit caveat, if the store is unavailable. Each record carries a canonical-JSON SHA-256 integrity hash a third party can re-derive offline. How verification works
preflight_action v0 public draft
A versioned, reviewable vocabulary for composing read-only preflight evidence into an agent or treasury workflow. Publication does not make unimplemented action types callable. Spec, schema, and changelog
Coinbase AgentKit provider
Published on npm as @routescore/[email protected] with four read-only actions: routescore_check_swap, routescore_quote_check, routescore_policy_evaluate, and routescore_get_preflight_record. Install with npm install @routescore/agentkit. The package is prepared for a separate upstream contribution to coinbase/agentkit; npm publication does not imply Coinbase endorsement or upstream inclusion.
Public read-only surfaces
Public snapshot list
Lists available public detector and benchmark snapshots with freshness metadata.
Public snapshot payload
Returns an allowlisted public scanner payload, such as mev_loss_index, stablecoins, bridges, or oracle_map.
Supported-token prices
Prices only the supported launch token universe. This is not a universal token pricing API.
Route score quote race
Returns supported route quote comparison data with rate limits and short cache windows.
User-owned exports
Account export
Authenticated JSON export containing account, saved wallets, and hash-recorded journal records.
Journal CSV export
Authenticated CSV export for durable hash-recorded journal entries.
Detector refresh requests
Power-tier queue for supported public detector snapshots. Not a contractual freshness guarantee.
{
"generated_at": "2026-06-19T00:00:00.000Z",
"source": "routescore",
"coverage": {
"mode": "supported",
"not_included": ["unsupported tokens", "wallet-specific MEV reconstruction"]
},
"scenario": {
"kind": "stress_test",
"modeled": true,
"point_in_time": true,
"not_investment_advice": true
}
}- Copy scenario JSON into a notebook or internal memo.
- Export journal and wallet records for personal archives.
- Build public detector dashboards from snapshot endpoints.
- Run pre-sign swap checks before an agent or wallet asks for a signature (free agent tier: 100/day).
- Call modeled premium estimates & what-if from your own agents via the API / MCP (Power).
- Retrieve persisted preflight evidence records and re-verify their hashes offline.
- Teach an agent the preflight + relay discipline with the routescore-ai skills.