Agent-First Kit

Command-line tools built for agents, not terminals. Every tool reports results as structured JSON events — typed values, machine-readable error codes, and timing — so an agent never has to parse human prose. Tools that agents can read.

Four Agent-First Tools

One shared naming convention, and CLI tools that follow it — so every field, flag, and error code is self-describing.

agent-first-data

The naming convention and JSONL output protocol. Encode units and semantics in the field name itself, so agents parse structured data without an external schema.

Agent-First Data →

agent-first-http

An HTTP client that reports every request and response as structured events on stdout — status, headers, timing, and typed errors.

Agent-First HTTP →

agent-first-pay

Payment tooling that emits machine-readable events for every operation, with Agent-First Data suffix conventions throughout.

Agent-First Pay →

agent-first-psql

A PostgreSQL client that reports rows, timing, and SQLSTATE failures as machine-readable events instead of terminal prose.

Agent-First PSQL →