OpenHumanvsOrkas
Generated from the two spec rows. Green marks the better value where a spec has a clear direction. Everything else is just different.
OpenHuman
Tiny Humans · Agent harness
OSSMCP
Panel
5.53 spec wins
- Reliability
- 4.7
- Usefulness
- 5.5
- Cost
- 7.3
- Longevity
- 4.5
“Seventeen messaging channels, which is sixteen additional ways to be told the task failed.”
Orkas
Orkas-AI · Agent harness
OSS
Panel
5.02 spec wins
- Reliability
- 3.7
- Usefulness
- 5.5
- Cost
- 6.3
- Longevity
- 4.7
“Its multi-agent team can execute terminal commands but lacks a Docker sandbox for isolation.”
Spec by spec
| Spec | OpenHuman | Orkas |
|---|---|---|
| Architecture | ||
| Category | Agent harness | Agent harness |
| Runs | local | local |
| Platforms | macos, linux, windows | macos, windows, linux |
| Context window | not documented | not documented |
| Protocols | ||
| MCP client | Yes | No |
| MCP server | No | No |
| Capabilities | ||
| Runs terminal commands | Yes | Yes |
| Multi-file edits | No | Yes |
| Git operations | No | Yes |
| Browser control | Yes | No |
| Sandboxed execution | No | No |
| Multi-agent orchestration | Yes | Yes |
| Headless / CI mode | No | No |
| Models | ||
| Backbone | any | Claude, OpenAI, Gemini, DeepSeek, Kimi, GLM, Qwen, MiniMax |
| Bring your own model | Yes | Yes |
| Local models | Yes | Yes |
| Cost | ||
| Pricing model | mixed | mixed |
| Starts at | n/a | n/a |
| Free tier | Yes | Yes |
| Bring your own key | Yes | Yes |
| Openness | ||
| Open source | Yes | Yes |
| License | GPL-3.0 | MIT |
| GitHub stars | 40,434 | 2,144 |
Which one would each critic pick
| Critic | OpenHuman | Orkas | Pick |
|---|---|---|---|
| El Juez | — | — | not enough reviews |
| El Amigo | 5.5 | 5.3 | OpenHuman — Pick OpenHuman if you want an agent that reaches you in the apps you already use all day; pick Letta if you want the memory without also inviting it into your messages. |
| El Crítico | 4.8 | 5.5 | Orkas — Orkas runs agents with terminal and git access on your local machine without a sandbox, a significant security risk. |
| El Profesor | 6.0 | 5.3 | OpenHuman — Memory is compressed into Markdown trees held in SQLite, and sub-agent fleets execute on checkpointed graphs, so both retention and resumption are explicit mechanisms rather than prompt tricks. |
| La Inversora | 5.3 | 5.5 | Orkas — A powerful local-first agent orchestrator, but its business model is a head-scratcher; use it, but don't bet your company on its longevity without a clearer revenue signal. |
| La Jefa | 3.8 | 3.0 | OpenHuman — This reads employee communications across seventeen personal channels, which is a data protection question before it is a tooling question, and the answer is no. |
| El Hacker | 7.8 | 5.8 | OpenHuman — GPL-3.0 means a fork stays open, it takes a local Ollama endpoint, it speaks MCP as a client, and there is no published install command, so I clone it anyway. |
Picks are derived from each critic's own scores. Humans vote on matchups on the duels page.