your llm knows gh for github,aws for aws.oto covers the long tail.
one unified cli for every saas that never shipped one. notion, linkedin, pappers, folk, hunter, google workspace — fifteen connectors and counting, ready to call from bash by any agent.
№ 01 — the argument
why cli over mcp.
mcp injects each api's full schema into the model's context. at the scale of an autonomous agent, cost and reliability collapse.
read the full benchmarks→public 2025 benchmarks — native composability via pipes, jq, xargs. works with claude code, codex, or any agentic framework.
№ 02 — anatomy
three files. one connector.
a cli command
auto-discovered. drop a file in commands/, it shows up in oto --help. json output, parsable by any llm.
a python client
importable from any script. the code the cli uses is the code you use.
a skill.md
the connector's manual, written for the llm. a few hundred tokens. enabled via oto skills enable.
№ 03 — catalog
what oto brings to the table.
fifteen connectors in production, grouped by access mode.
№ 04 — get started
three lines to get going.
oto is built by alexis laporte, independent ai engineer. open source · mit license.