Terminal-native coding agent
Weave code with an agent
you can actually inspect
Loom is a coding agent for your terminal. Point it at any
OpenAI-compatible endpoint, hand it a goal, and it reads, edits, runs
and verifies — writing
every state change to an append-only journal you can replay, rewind and audit.
$
cargo install --path crates/loon-cli --force
Or build from source: cargo build --release → ./target/release/loom.
The default provider is a deterministic mock, so the first run needs no credentials.
loom — ~/projects/api
╭─ Loom · ultra · auto
│
❯ fix the token parser and prove it
▸ read crates/loon-config/src/tokens.rs
▸ edit tokens.rs (+18 −4)
+ "1m" => TokenSpec::Tokens(1_000_000),
+ "100k" => TokenSpec::Tokens(100_000),
- parse(s).unwrap()
▸ bash cargo test -p loon-config tokens
✓ 14 passed · 0 failed · 0.31s
context ~41.2k / 200k (21%)
── verified · journal seq 1,204 · replayable