Score breakdown
Popularity is tracked separately. Support, ads, sponsorships, and tips never affect these signals.
Why it matters
Useful for AI-coding-agent teams and platform teams that want a maintained MIT open-source agent harness that treats the model itself as a plugin rather than the center of a vertically integrated stack (Claude Code / Codex / OpenCode), for developers who want to inspect / edit / test repositories, run shell and development tools, plan, delegate to subagents, and enforce configurable permission
Where this stands now
deepseek-ai/deepseek-harness ranks #10 of 33 tracked Coding Agents items by composite score (8.2 against a section median of 8.0). The section currently carries 20 Gold, 10 Silver, 3 Bronze. RepoRadar has retained observations for this record since 2026-08-14 (30 days in the current window). Signal extremes versus the section: momentum at the 86th percentile; novelty at the 80th percentile.
Who should use it
Who should skip it
Skip deepseek-ai/deepseek-harness unless the captured evidence suggests it solves a problem you are actively working on.
About this signal
deepseek-ai/deepseek-harness is tracked by RepoRadar as a framework in the Coding Agents section. First seen 2026-08-14; the source record was last checked on 2026-08-14. The current verdict is 'try now' with a Gold tier and moderate setup difficulty. deepseek-ai/deepseek-harness leads on workflow potential (9.7) and momentum (9.0); its lowest signal is setup ease (6.8), so factor that in before investing setup time. This page summarizes the evidence RepoRadar captured from https://github.com/deepseek-ai/deepseek-harness.
How this item is evaluated
The deepseek-ai/deepseek-harness record combines a 8.2/10 composite score with separate popularity (100.0), risk (low), and setup (moderate) signals. See the scoring methodology for the current weights and evidence definitions.
Putting this into practice? Read How to vet an AI agent or MCP server before you wire it in for the checklist behind this score.
Risk explanation
explicit developer-preview status with planned compatibility-breaking changes -- pin the version and review the plugin contract before adopting in production; the model endpoint is a plugin, so the same agent loop runs against any compatible provider -- but provider-specific plugins may diverge on tool-calling conventions, so smoke-test against your chosen endpoint before relying on it; the Web UI binds to 127.0.0.1:3080 by default -- if you ever expose it beyond localhost, configure auth + TLS rather than relying on the loopback-only default; the npm package @deepseek-ai/dsh is the install entry point; pin the version explicitly for reproducibility.