cargo-terse

Concise cargo output for AI-assisted Rust development.

cargo-terse wraps cargo build, check, test, and clippy with condensed output optimized for AI agent consumption. Fewer tokens, same information.

Install

From crates.io

cargo install cargo-terse

Pre-built binaries (via cargo-binstall)

cargo binstall cargo-terse

From source

git clone https://github.com/dandanlen/cargo-terse.git
cd cargo-terse
cargo install --path .

Usage

cargo terse check           # concise check output
cargo terse test             # test results without noise
cargo terse clippy           # terse clippy warnings
cargo terse                  # defaults to check

Setup

cargo terse setup                  # auto-detect and configure agent instruction files
cargo terse setup --global         # configure ~/.claude/CLAUDE.md
cargo terse setup --agent cursor   # configure .cursorrules specifically

Auto-detects existing agent config files (CLAUDE.md, AGENTS.md, .cursorrules, .github/copilot-instructions.md) and appends usage instructions. Idempotent — won't duplicate if already configured.

Output formats

cargo terse --format plain check   # default: one-line-per-diagnostic
cargo terse --format json check    # JSONL output
cargo terse --format toon check    # TOON (Token-Oriented Object Notation)

Verbosity

cargo terse check          # terse (default): one-liner per diagnostic
cargo terse -v check       # verbose: includes code span
cargo terse -vv check      # very verbose: full rustc rendered output

Drill-down

Each diagnostic gets an ID. Use detail to expand it:

cargo terse check          # shows: W1 warning[unused_variables] src/lib.rs:4:9 ...
cargo terse detail W1      # shows full rendered diagnostic for W1

Configuration for AI agents

Add to your project's CLAUDE.md (or equivalent AI agent instructions):

Use `cargo terse` instead of `cargo` for build, check, test, and clippy commands.
Examples:
- `cargo terse check` instead of `cargo check`
- `cargo terse test` instead of `cargo test`
- `cargo terse clippy` instead of `cargo clippy`

When a diagnostic needs more context, use `cargo terse detail <ID>` to expand it.

Example output

Default (terse):

W1 warning[unused_variables] src/lib.rs:4:9 unused variable `extra`
ok (check) 1 warning 0.1s

JSON:

{"id":"W1","level":"warning","code":"unused_variables","file":"src/lib.rs","line":4,"col":9,"message":"unused variable `extra`"}
{"summary":true,"command":"check","status":"ok","warnings":1,"errors":0,"elapsed_secs":0.1}

Test output:

ok (test) 36 passed, 0 failed 0.5s

With failures:

F1 FAILED tests::parse_config_missing_field
F2 FAILED tests::handler_timeout
test result: FAILED. 2 passed; 2 failed; 0 ignored 1.2s

Why not --message-format=short?

cargo check --message-format=short has existed since 2018 and reduces diagnostic verbosity. cargo-terse builds beyond it:

• Stable drill-down IDs (W1, E1, F1) let you revisit any diagnostic with cargo terse detail <ID> without re-running the build
• Unified output across check, test, and clippy in a single consistent format
• Multiple output formats (plain, JSON, TOON) suited to different consumers
• Pairs well with RTK for additional token savings on broader cargo output

How it works

cargo-terse spawns cargo with --message-format=json (rather than the simpler --message-format=short), parses the JSON stream in real-time, and re-renders diagnostics in a condensed format. The JSON stream provides enough detail to assign stable IDs and support drill-down. Test results are parsed from stdout text output. A cache file (target/.terse-cache.json) enables the detail drill-down command.

Flags

All other flags pass through to cargo.

License

MIT