StackMap
Subscribe

mini-swe-agent vs OpenManus

The 100-line agent from the SWE-bench team: >74% on SWE-bench Verified with no tools but bash, no config sprawl — the reference minimal harness, adopted by Meta, NVIDIA and Ramp. — versus — Open-source general autonomous agent from the MetaGPT team — the 'Manus without an invite code': browsing, tool use and multi-step task execution from a simple Python core.

The curated verdict

Both are open, minimal-core autonomous agents you can actually read. OpenManus goes general (browsing, tool use, multi-step tasks); mini-swe-agent goes narrow and measurable — software engineering with bash only, scored on SWE-bench.

mini-swe-agentOpenManus
Stars6.0k58k
Forks83110k
LanguagePythonPython
LicenseMITMIT
Last activityyesterday5 months ago
Topicscoding, agentsagents
Curated connections23

mini-swe-agent — the curator's take

The existence proof that most harness complexity is optional: the team that built SWE-bench and SWE-agent asked what a 100x simpler agent loses — the answer is almost nothing (>74% Verified), which is why it became the standard baseline harness for benchmarking models (Ramp's SWE-bench, DeepSWE — where it beats Claude Code and Codex as a harness). Read it to understand agents; use it to evaluate models fairly. NOT a daily driver: no MCP, no skills, no IDE plumbing — by design. If you're choosing a tool to ship features with, this is the control group, not the product.

OpenManus — the curator's take

The fastest way to feel what a general autonomous agent does: clone, add a model key, hand it a task — the 3-hour-prototype energy that earned 57k stars keeps the codebase small enough to actually read, which makes it a great learning skeleton. But look at the commit graph before betting on it: activity has been quiet for months while the team's focus moved on (OpenManus-RL and beyond), so treat it as a reference implementation, NOT a maintained production framework — for durable agent infrastructure reach for an actively developed harness instead.