herdrvsMateClaw
Generated from the two spec rows. Green marks the better value where a spec has a clear direction. Everything else is just different.
herdr
herdr · Agent harness
OSS
Panel
6.31 spec wins
- Reliability
- 6.2
- Usefulness
- 5.7
- Cost
- 7.5
- Longevity
- 6.0
“A tmux for agents whose detach key is ctrl+b q, so twenty years of muscle memory finally counts as onboarding.”
MateClaw
MateOS · Agent harness
OSSMCP
Panel
6.35 spec wins
- Reliability
- 5.7
- Usefulness
- 6.2
- Cost
- 8.0
- Longevity
- 5.5
“The desktop app ships with its own bundled JRE 21, because the one thing worse than asking is finding out which Java they had.”
Spec by spec
| Spec | herdr | MateClaw |
|---|---|---|
| Architecture | ||
| Category | Agent harness | Agent harness |
| Runs | local | local |
| Platforms | macos, linux, windows | macos, linux, windows, web |
| Context window | not documented | not documented |
| Protocols | ||
| MCP client | No | Yes |
| MCP server | No | No |
| Capabilities | ||
| Runs terminal commands | Yes | YesSensitive tool actions are approval-gated behind a Tool Guard; the coding work itself runs in the DeepSeek Harness runtime when that backend is selected. |
| Multi-file edits | No | YesDelivered through the agent runtime in use rather than by MateClaw's own tools. |
| Git operations | No | No |
| Browser control | No | No |
| Sandboxed execution | No | No |
| Multi-agent orchestration | Yes | YesThe native runtime supports team runs and A2A connects governed employees across systems. |
| Headless / CI mode | No | No |
| Models | ||
| Backbone | Claude Code, Codex, Cursor, OpenCode, Grok | DashScope, OpenAI, Anthropic, Gemini, DeepSeek, Kimi, Ollama, LM Studio, MLX |
| Bring your own model | No | Yes |
| Local models | No | Yes |
| Cost | ||
| Pricing model | byok | byok |
| Starts at | $0/mo | $0/mo |
| Free tier | Yes | Yes |
| Bring your own key | No | Yes |
| Openness | ||
| Open source | Yes | Yes |
| License | Apache-2.0 | Apache-2.0 |
| GitHub stars | 41,859 | 1,078 |
Which one would each critic pick
| Critic | herdr | MateClaw | Pick |
|---|---|---|---|
| El Juez | — | — | not enough reviews |
| El Amigo | 7.0 | 6.0 | herdr — Pick herdr if your Claude Code, Codex, Cursor or OpenCode sessions die when you close the laptop; pick Claude Squad if what you actually need is a worktree per task. |
| El Crítico | 6.0 | 5.5 | herdr — Agents can spawn panes and prompt other agents on one host with no sandbox and no worktrees, so a chain of agents shares one working tree and one set of permissions. |
| El Profesor | 6.8 | 7.0 | MateClaw — Persistent goals are checkpointed in the database with leases, attempts and cooldowns, and a supervisor reconciles them after a restart. That vocabulary comes from job scheduling, correctly. |
| La Inversora | 4.8 | 5.3 | MateClaw — First release in April 2026, published on Gitee with 499 stars there against 1,078 on the mirror, no company behind the name and nothing metered. This is very early. |
| La Jefa | 5.5 | 6.5 | MateClaw — Multi-user workspaces, a full audit trail, approval gates on sensitive actions and Actuator health endpoints, self-hosted as one JAR with no per-seat charge at all. |
| El Hacker | 8.0 | 7.8 | herdr — Apache-2.0, one binary via brew or mise, a socket API I can script from anything, no MCP and no model settings because it owns terminals, not agents. |
Picks are derived from each critic's own scores. Humans vote on matchups on the duels page.