Use txcast with AI agents

txcast exposes the same organisation-scoped commands through:

• remote Streamable HTTP MCP at https://api.txcast.com/mcp; and
• JSON REST under https://api.txcast.com/agent/v1.

Use MCP with Codex, Claude Code, or another compatible host. Use REST when your automation already works with HTTP or OpenAPI. Both use the same credential, validation, authorization, account grants, idempotency, and audit provenance.

Safe starting path

1. Create a short-lived, selected agent credential.
2. Start with read-only or authoring scopes.
3. Connect Codex or Claude Code.
4. Ask the agent to inspect context and saved contracts before writing.
5. Review every sensitive tool confirmation.
6. Sign or broadcast only through a separate wallet capability.

What agents can do

Depending on scopes, agents can inspect or register contracts, author complete workflow drafts, publish revisions, manage runs, prepare Safe/EOA handoffs, submit Safe signatures, and record transaction hashes.

They cannot hold keys, sign, broadcast, administer the organisation, replace/delete ABIs, delete workflows, or mutate controlled accounts through v1.

Caution

Client confirmation prompts can be disabled and are not a security boundary. The durable boundaries are credential scope, controlled-account grants, live organisation authority, optimistic concurrency, cryptographic verification, and wallet confirmation.

Provenance

Every explicit agent mutation produces one organisation audit event with source, credential snapshot, and client-reported agent name/version. Reads are not audited. Replaying the same idempotency key returns the original result and audit.