Decision Workspace
agent-harness vs claude-agent-rs vs agent-session
Side-by-side comparison of Rust crates
46
agent-harness
experimentalv0.5.3
Drive LLM coding agents — Claude Code, OpenAI Codex, and local Ollama / OpenAI-compatible models — from Rust behind one trait, with a normalized streaming event vocabulary. Bring your own agent too.
49
claude-agent-rs
growingv1.1.0
Lightweight AI coding agent in Rust — runs Claude on your Max subscription for $0. 8 built-in tools, MCP client, hooks, skills with frontmatter, sessions, sub-agents, streaming.
65
agent-session
experimentalv0.5.17
Portable session IR, parsers, discovery, and process matching for local AI coding-agent transcripts.
Core Metrics
| agent-harness | claude-agent-rs | agent-session | |
|---|---|---|---|
| Health Score | 46 | 49 | 65 |
| Total Downloads | 512 | 271 | 7.0K |
| 30d Downloads | 0 | 0 | 0 |
| Dependents | 0 | 0 | 246 |
| Releases | 19 | 12 | 79 |
| Last Updated | 34d ago | 191d ago | 5d ago |
| Age | 4m | 7m | 3m |
Health Breakdown
agent-harness
Maintenance
17
Quality
8
Community
6
Popularity
3
Documentation
12
claude-agent-rs
Maintenance
11
Quality
16
Community
6
Popularity
3
Documentation
13
agent-session
Maintenance
19
Quality
13
Community
13
Popularity
5
Documentation
15
Technical Details
| agent-harness | claude-agent-rs | agent-session | |
|---|---|---|---|
| Version | 0.5.3 | 1.1.0 | 0.5.17 |
| Stable (≥1.0) | ✗ No | ✓ Yes | ✗ No |
| License | MIT OR Apache-2.0 | MIT | MIT |
| Dependencies | 17 | 15 | 7 |
| Crate Size | 305KB | 67KB | 56KB |
| Features | 6 | 0 | 0 |
| Yanked % | 26.3% | 0.0% | 0.0% |
| Edition | 2021 | 2021 | 2024 |
| MSRV | 1.74 | — | 1.87 |
| Owners | 1 | 1 | 1 |
Links
Quick Verdict
- •agent-session leads with a health score of 65/100, but none of the options score above 80.
- •agent-session is depended on by 246 crates — strongest ecosystem trust.
- •agent-harness, agent-session are pre-1.0 — API may change.