docker-agent vs penguin-harness
Two of the top projects in Agent and MCP starters, side by side: who leads on each score part and who each one suits, then setup, license, activity and what each review found.
docker-agent
Docker CLI plugin that runs multi-agent teams defined in YAML
penguin-harness
Agent framework that evolves an agent's prompts, skills and tools against benchmarks
Verdict
docker-agent ranks higher, 62 to 61. Of the 5 score parts we measure for both, docker-agent leads on 2, penguin-harness on 1 and 2 are even.
- Adoptiondocker-agent leads by 15 points, 64 to 49: a wide gap.
- FreshnessEven at 100.
- MaintenanceEven at 93.
- Easy to runpenguin-harness leads by 16 points, 33 to 17: a wide gap.
- Agent-readydocker-agent leads by 30 points, 30 to 0: a wide gap.
Pick docker-agent if you are
Developers building multi-agent setups from YAML with MCP tools
- Agents, tools and models declared in versionable YAML, no code required
- Works with local, remote or Docker-based MCP servers as toolsets
Pick penguin-harness if you are
Engineers tuning agent prompts and skills, or self-hosting agents
- Five RSI algorithms run on the same benchmarks, scoreboard and per-version snapshots
- Desktop app, Web App, CLI and SDK share one engine and the ~/.penguin/data directory
| What we compare | docker-agent | penguin-harness |
|---|---|---|
| Score parts, out of 100 | ||
| Adoption | 64, popular | 49, known |
| Freshness | 100, active | 100, active |
| Maintenance | 93, healthy | 93, healthy |
| Easy to run | 17, hard | 33, some setup |
| Agent-ready | 30, minimal | 0, none |
| Facts from GitHub and the README | ||
| Stars | 4.6k | 2.9k |
| License | Apache-2.0 (permissive) | Apache-2.0 (permissive) |
| Last commit | Oct 2026 | Oct 2026 |
| Last release | Oct 2026 | Oct 2026 |
| Language | Go | TypeScript |
| Docker | Yes | Yes |
| GPU | Optional | Not needed |
| arm64 or Apple Silicon | Mentioned | Mentioned |
docker-agent
docker-agent is a Go CLI plugin (`docker agent run agent.yaml`) that runs agents declared in YAML: model, instructions, and toolsets, including any MCP server. Agents can delegate to other agents, and configs can be pushed to and pulled from OCI registries. It supports OpenAI, Anthropic, Gemini, AWS Bedrock, Mistral, xAI and Docker Model Runner, plus built-in RAG with BM25, embeddings and reranking.
Who it is for: Developers building multi-agent setups from YAML with MCP tools
Strengths
- Agents, tools and models declared in versionable YAML, no code required
- Works with local, remote or Docker-based MCP servers as toolsets
- Agent configs can be shared through any OCI registry
- Many providers supported, plus local models via Docker Model Runner
Weaknesses
- Collects anonymous usage telemetry; opt-out details are only in the docs
- Needs a cloud provider API key or Docker Model Runner to do anything
- Configuration is YAML-centric; custom logic needs MCP tools rather than code hooks
- RAM and hardware requirements are not documented in the README
- Go, OpenAI, Anthropic, Gemini, AWS Bedrock
- Needs Docker CLI (plugin), LLM provider API key or Docker Model Runner, MCP servers (optional), OCI registry (optional)
- Docker
penguin-harness
PenguinHarness runs agents with a small toolset and keeps the model fixed while it evolves the harness (prompts, Skills, tools, hooks) over scored rounds. Five RSI algorithms (default, OPRO, APE, ACE, AWM) ship as plugins and run on built-in Harbor benchmarks in Docker, with each candidate snapshotted as a version. It ships as a desktop app, Web App, CLI and TypeScript SDK, and can also generate agent apps from a prompt.
Who it is for: Engineers tuning agent prompts and skills, or self-hosting agents
Strengths
- Five RSI algorithms run on the same benchmarks, scoreboard and per-version snapshots
- Desktop app, Web App, CLI and SDK share one engine and the ~/.penguin/data directory
- Works with many providers plus custom OpenAI-protocol endpoints, including local models
- Agents can run on remote machines over ssh and drive Chrome through an extension
Weaknesses
- Benchmarks run as Harbor tasks in Docker, so self-evolution needs Docker
- Cost and accuracy claims come from the project's own suite, not independent runs
- RAM and disk requirements are not documented
- Company-level self-evolution and OpenShell integration are still unchecked roadmap items
- TypeScript, OpenAI, Anthropic, Google Gemini, OpenRouter
- Needs Node >= 24 (npm install), Docker (benchmarks), Model provider API key
- Docker
- env example file
More in Agent and MCP starters
Building with these?
We help teams pick an open-source AI stack and run it in production. Client work never changes a rank.