Charter declares agent API tools as Pydantic schemas, auto-handling request construction, auth, wire format, and response trimming instead of hand-code.
Yes, Charter is open source under the Apache-2.0 license.
Charter is free to use.
Yes, Charter can be self-hosted (the source is available under the Apache-2.0 license).
Charter is a Python library that lets you declare agent API tools as Pydantic schemas instead of hand-implementing them, auto-handling request construction, auth, wire format, and response trimming. Open source: yes (Apache-2.0); self-hostable; pricing free. It runs in-process with no telemetry, launched 2026-09-28, and has 1 star.
Charter replaces the usual pile of hand-written tool functions with a declaration. You describe each API tool once as a Pydantic schema, and Charter does the plumbing: it constructs the request, handles authentication, deals with the wire format, and trims the response down before it reaches the model — so the agent both calls the API correctly and does not drown in raw payload. It is a plain Python library that runs in-process, with no telemetry phoning home. Open source: yes (Apache-2.0); self-hostable; pricing free.
A GTM agent's usefulness is mostly its tools — CRM writes, enrichment lookups, GraphQL queries — and each hand-built tool is code to get right and bytes of response to feed the model on every call. Charter attacks both: declare the tool as a schema and let the library enforce correct, authorized requests and trim the response, which directly cuts malformed calls and token spend. Open source: yes (Apache-2.0); self-hostable; pricing free. The honest read: the declarative-tool idea is sound and the maker's 534-run numbers (fewer malformed/unauthorized calls, ~4x fewer forwarded bytes) point the right way, but this is a Python-only, day-one project at 1 star — trial it on one integration and confirm the lift before standardizing your tool layer on it.