How an AI agent can compare USDC bridge fees before it executes
A preflight step for agents that move stablecoins between blockchains: compare the routes at one chain state, keep only a provable saving, and verify the delivered amount after execution. Free API, no key.
The problem
An agent that bridges USDC usually calls one bridge or one aggregator and signs what it gets. That quote may include fees the agent never compares, and quotes from different providers fetched at different moments are not comparable. Over many transfers the difference adds up, and nobody can show the principal that the route chosen was the cheap one.
A preflight that compares and proves
- 1. Before signing, the agent declares the route it would use anyway (its baseline).
- 2. A preflight reads every candidate route at the same anchor block and computes what each delivers after fees and gas.
- 3. It answers with the cheapest route only when the saving holds as a lower bound; otherwise it says the baseline is already the best.
- 4. After execution, the delivered amount is checked on chain, so the saving is a measured number, not a quote.
Recta is that preflight, as an HTTP API, an MCP server and npm packages. The evaluation is free; when the agent executes the cheaper route Recta charges exactly 5 % of the saving it verified, and nothing when there is none.
Wire it into an agent
- MCP: add the MCP server (Streamable HTTP), or run
npx @recta/mcplocally; its free tools check the corridor and run an unsigned trial without a key. - TypeScript:
npm install @recta/client. - Plain HTTP: llms.txt and OpenAPI.
Check your own transfer, free, without a wallet
Recta is an API for AI agents (and the people who build them) that compares the routes for a cross-chain stablecoin transfer at the same chain state and returns the cheapest one it can prove, measured against the route you would have used anyway. The evaluation is free. If you then execute the cheaper route, Recta charges exactly 5 % of the saving it verifies on chain; no saving, no charge.
- 1. Is the corridor covered, and which routes are compared:
GET https://api.recta.build/v1/golden-corridor/fit?source_network=ethereum&source_asset=USDC&destination_network=arbitrum&destination_asset=USDC - 2. The request body for your amount and the route you use today:
GET https://api.recta.build/v1/golden-corridor/request-template?source_network=ethereum&source_asset=USDC&destination_network=arbitrum&destination_asset=USDC&amount=5000&baseline_provider=lifi - 3. Send that body, with your addresses and gas ceilings filled in, to
POST https://api.recta.build/v1/golden-corridor/trial-evaluation. No signature and no wallet: it answers whether a saving exists, its lower bound and what 5 % of it would be (3 trials a day per IP address). - 4. To get the route itself, sign the same body and send it to the signed evaluation (see llms.txt), or use the npm packages
@recta/clientor@recta/mcp, or the MCP server.
How often Recta has found a saving on this corridor, per route you could declare: saving odds.