Open sourcePython · ≈50k stars · Apache-2.0
Aider
AI pair programming in your terminal.
Aider is the git-aware CLI pair programmer that predates the coding-agent wave and still defines it for many: in-repo edits, dozens of models, repo maps, and now MCP support so agents see version control and tools. The harness that proved developers would live in the loop.
Pros & cons
Each point cites its sources and says what kind of evidence backs it. A second reviewer tried to refute every claim before it was kept.
Pros
- User reportsCredited by users as the tool that defined terminal coding agents, with plan-style modes before Claude Code.[1][2][3]
- User reportsUsers praise architect mode, where a pricier model plans and a cheaper one edits, as cost-effective.[1][2][3]
- User reportsWorks with many models, including local ones via Ollama, and can be scripted.[1][2]
- MetadataApache-2.0 and git-native: it commits its changes, so they are easy to review and revert.[1]
Cons
- User reportsLess autonomous: users note it lacked an automatic write-run-fix loop and defaulted to limited multi-turn recursion.[1][2]
- User reportsSome users find it feels like chat in a terminal, with nagging prompts and 'too much busywork'.[1][2]
- MetadataDevelopment has slowed: the last release was v0.86.0 (Aug 2025), the last push was May 2026, and 1,408 issues are open. Users report moving to newer harnesses.[1][2][3]
- User reportsUsers report quality dropping in long conversations and heavy token use (2M tokens a day for one user).[1][2]
- Best for
- Developers who want a controlled, git-native pair programmer with explicit file context and any model.
- Watch out for
- Slowing development and a less agentic loop than current harnesses.
Assessed Oct 29, 2026 · adversarially reviewedSuggest a correction
People
Founder
Creator of Aider
Open harnesses · Evals & benchmarks
Activity
- Last push
- · 6mo ago
- Latest release
- v0.86.0
· 1y ago - Last 90 days
- 0 releases
v0.86.0Aider v0.86.0
· 1y agoAdded support for all GPT-5 models. Added support for Grok-4 via xai/grok-4 and openrouter/x-ai/grok-4 model names. Added support for gemini/gemini-2.5-flash-lite-preview-06-17 model, by Tamir Zahavi-Brunner. /clear now prints “All chat history cleared.” so you know it worked,…
v0.85.0Aider v0.85.0
· 1y agoNew Models Added support for new Gemini models including gemini-2.5-pro, gemini-2.5-flash, and gemini-2.5-pro-preview-06-05 with thinking tokens support. Updated model aliases: flash now points to gemini-2.5-flash and gemini now points to gemini-2.5-pro. Support for Responses…
v0.84.0Aider v0.84.0
· 1y agoModel support & provider updates Added support for new Claude models including the Sonnet 4 and Opus 4 series (e.g., claude-sonnet-4-20250514, claude-opus-4-20250514) across various providers. The default sonnet and opus aliases were updated to these newer versions. Added…
v0.83.0Aider v0.83.0
· 1y agoAdded support for gemini-2.5-pro-preview-05-06 models. Added support for qwen3-235b models. Added repo-map support for OCaml and OCaml interface files, by Andrey Popp. Added a spinner animation while waiting for the LLM to start streaming its response. Updated the spinner…
v0.82.0Aider v0.82.0
· 1y agoSupport for GPT 4.1, mini and nano. Improved support for using architect mode with Gemini 2.5 Pro. Add support for xai/grok-3-beta, xai/grok-3-mini-beta, openrouter/x-ai/grok-3-beta, openrouter/x-ai/grok-3-mini-beta, and openrouter/openrouter/optimus-alpha models. Added support…
Checked against GitHub on Nov 21, 2026. Release notes are shortened to plain text.
In research (9)
- Evaluation-Free Recursive Self-Improvement for Coding AgentsVerdict: Useful
Nov 16, 2026 · Source of the Polyglot benchmark used for transfer evaluation.
- RepoTrace: Improving Requirement Tracing in Multi-Round Agent InteractionsVerdict: Incremental
Oct 26, 2026 · Aider's graph-ranked repo map is the obvious prior art and a missing baseline.
Oct 26, 2026 · Aider's dependency-graph repo map is close prior art for budgeted repository context.
- CTXAgent - Self-Controlled Context Management of LLM AgentsVerdict: Useful
Oct 25, 2026 · Coding exercises come from Aider's polyglot benchmark.
- NameTrap: Steering AI Coding Agents Toward Security Vulnerabilities via Variable NamesVerdict: Useful
Oct 25, 2026 · Most vulnerable agent in the study (56.0% overall ASR).
- ProductWebBench: Evaluating Website Continuity Beyond Task CompletionVerdict: Important
Oct 22, 2026 · Evaluated scaffold on both backbones.
Showing the 6 newest of 9 papers.
Neighbours on the workbench
Shell · ≈329k stars · MIT
mattpocock/skills
Small, composable agent skills for real engineering work.
Maintained by Matt Pocock (AI Hero)
MDX · ≈323k stars · MIT
obra/superpowers
An agentic skills framework and software methodology that works.
Maintained by Jesse Vincent (obra)
JavaScript · ≈308k stars · MIT
affaan-m/ECC
Agent harness performance optimization system: skills, instincts, memory and security for coding agents.
Maintained by Affaan Mustafa
Connections
Relations from the Harness Hub knowledge graph, as charted in the State of Harness reports.
