Hi @bighadj22 - I'm building AstraNL, a non-custodial coordination protocol for AI agents (Netherlands, KvK 88449335). Found codflow while looking for MCP servers where an agent can actually commit money or a booking on someone's behalf - purchase is exactly that surface.
One thing MCP itself doesn't cover: when an agent calls such a tool, your server can't tell whether the human behind the agent really authorised that spend (scope, ceiling, time window), and nothing durable remains afterwards for a dispute.
We shipped a two-call handshake for that:
POST https://astranl.com/api/amh/verify before the action - returns ok / ceiling / window for the agent's mandatePOST https://astranl.com/api/amh/seal after it - a sealed, publicly verifiable receipt
Free, no custody (AstraNL never touches payment), no SDK - two HTTP calls. Docs: https://astranl.com/api/amh
Not asking you to change defaults. Would an optional flag in codflow that calls verify before the consequential tool runs (when a mandate is present) be welcome? Happy to send a PR. Equally happy to hear this is noise for your use case - one reply either way and I won't follow up.
Hi @bighadj22 - I'm building AstraNL, a non-custodial coordination protocol for AI agents (Netherlands, KvK 88449335). Found codflow while looking for MCP servers where an agent can actually commit money or a booking on someone's behalf -
purchaseis exactly that surface.One thing MCP itself doesn't cover: when an agent calls such a tool, your server can't tell whether the human behind the agent really authorised that spend (scope, ceiling, time window), and nothing durable remains afterwards for a dispute.
We shipped a two-call handshake for that:
POST https://astranl.com/api/amh/verifybefore the action - returns ok / ceiling / window for the agent's mandatePOST https://astranl.com/api/amh/sealafter it - a sealed, publicly verifiable receiptFree, no custody (AstraNL never touches payment), no SDK - two HTTP calls. Docs: https://astranl.com/api/amh
Not asking you to change defaults. Would an optional flag in codflow that calls
verifybefore the consequential tool runs (when a mandate is present) be welcome? Happy to send a PR. Equally happy to hear this is noise for your use case - one reply either way and I won't follow up.