Skip to content

zeta1999/simple-tools

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

2 Commits
 
 
 
 

Repository files navigation

simple tools

simple tools

Small, composable Rust libraries for building tools fast from a harness —
for local agents, or frontier agents with fewer tokens and faster models.

Rust 7 libraries local-first Le Harnais license

Umbrella for the simple-* libraries. Site: zeta1999.github.io/renoir42/simple-tools.html


Why

The brief for every library here is the same: keep it small enough that a harness — a coding agent that may be partly or entirely a local model, see Le Harnais — can wire it into a working tool in one sitting. No framework to learn, no plugin system, a surface you can read end-to-end and audit.

simple tools is the toolbox; the harness drives the model and the loop. Together with leanlift (proof automation) and a security toolkit, the goal is tools you can stand up quickly and trust.

The libraries

library status what it is
simple-ui released TUI widgets (tables, depth ladder, charts, agent-activity log) + a markdown engine with live IPC updates
simple-network released async transport with an optional post-quantum secure channel (ML-KEM + XChaCha20 + ML-DSA) and an optional Tor carrier
simple-secrets released post-quantum secret manager with advanced secret sharing, TOTP, and a written threat model
limited-shell coming soon a capability-scoped shell language; parse → type-check → schedule → execute under explicit limits (Lean 4 models)
simple-remote coming soon run a limited-shell session on a remote host over simple-network — same limits, no ambient SSH trust
simple-sandbox coming soon a sandbox for AI agents and the tools they run: contain side effects, mediate access, keep a trail
simple-backups coming soon content-addressed snapshots an agent can create and restore as part of a safe edit loop

The released libraries share a sibling, rust-secure-memory (secure allocation, zeroize, post-quantum KEM).

Design rules

  • Small over general. Each crate is meant to be read in an afternoon. No plugin systems, no config sprawl.
  • Capabilities, not trust. Execution is scoped by explicit limits (limited-shell), contained (simple-sandbox), and encrypted post-quantum on the wire (simple-network, simple-secrets).
  • Public means just the code. The -public mirrors carry the source that compiles — no internal history — with dependencies pointed at their own -public siblings.

License

MIT OR Apache-2.0

About

Tools and libraries for local agents (or frontier agents with less tokens/faster models)

Resources

Stars

0 stars

Watchers

0 watching

Forks

Releases

No releases published

Packages

 
 
 

Contributors