diff options
Diffstat (limited to 'content/posts')
| -rw-r--r-- | content/posts/coding-agent-specialised-tools.md | 165 |
1 files changed, 165 insertions, 0 deletions
diff --git a/content/posts/coding-agent-specialised-tools.md b/content/posts/coding-agent-specialised-tools.md new file mode 100644 index 0000000..32b62ee --- /dev/null +++ b/content/posts/coding-agent-specialised-tools.md @@ -0,0 +1,165 @@ +--- +title: "Can we do better than Read - Edit" +draft: false +date: 2026-06-14 +lastmod: +description: | + The landscape of coding agents tooling is fragmented, and there are no + definitive winners. +categories: + - LLM + - Coding Agents +--- + +Coding agents have agency - they can decide on their own +what files they need to read in order to understand a codebase. +Common harnesses such as Claude Code or Codex provide a +limited set of tools to perform Software Engineering tasks - +Read, Write, Edit. + +[Retrieval-Augmented Code Generation](#retrieval-augmented-code-generation-a-survey-with-focus-on-repository-level-approaches) +gives a good understanding of the alternative tools, splitting them in GRAPH vs +TEXT (BM25, Jaccard similarity). They also differentiate one-shot retrieval vs +dynamic exploration, where the agent has the opportunity to deepen its search +based on previous results. AST parsing can be used in both solutions. +Other approaches use sub-agents to explore and summarize. + +For pure analysis - answering a question such as "How does authentication +work in here?", a combination of graph database, AST parsing and +embedding models claim up to [50x saved input tokens](#code-review-graph). +This results in faster query answering, but it affects actual costs much +less, [dominated by output tokens](#grepai--benchmark-vs-grep--claude-code). + +In a 2024 paper for [SWE-agent](#swe-agent-agent-computer-interfaces-enable-automated-software-engineering), +we could already see a cost tradeoff - tool-based agents beat RAG or shell, +but at a tenfold cost increase. + +I couldn't find many benchmarks on editing. And to my knowledge, there are no +well-known benchmarks comparing all sorts of harnesses. Only the base CLIs +are compared (Claude Code vs Codex vs Opencode). The precise values I +get are always self-reported, and rarely compared. + +I also found a case study from [ManoMano and Serena](#manomano--serena--project-aegis), a tool +I tried for a week but didn't find so useful (or even used by Claude). +There was a slight cost increase, but it brought a huge quality boost. +Quick explorations are still better with Claude Code, but refactoring +is very fast and precise with Serena. + +I hoped I could wrap up this post with a definitive answer, but setting +up reproducible benchmarks to test all those tools is not as easy as I expected. +I think GRAPH-based solutions are more promising, because TEXT predates them +and were put aside by the big labs. I'll try to test some of them. + + +## Trending projects + +| Project | Stars | Capability | Technique | +|---|---|---|---| +| [codegraph](https://github.com/colbymchenry/codegraph) | 42.6k | Read, index | Graph DB | +| [GitNexus](https://github.com/abhigyanpatwari/GitNexus) | 40.8k | Read, Rename, Viz | Graph DB | +| [repomix](https://github.com/yamadashy/repomix) | 25.8k | Code compression? | Text packing | +| [serena](https://github.com/oraios/serena) | 24.8k | Read + Edit | AST / LSP | +| [code-review-graph](https://github.com/tirth8205/code-review-graph) | 17.2k | Read | Graph DB | +| [context-hub](https://github.com/andrewyng/context-hub) | 13.4k | Read | Embedding | +| [claude-context](https://github.com/zilliztech/claude-context) | 11.6k | Read | Embedding | +| [code2prompt](https://github.com/mufeedvh/code2prompt) | 7.4k | Read | Text packing | +| [semble](https://github.com/MinishLab/semble) | 4.6k | Read | BM25 + AST + Embedding | +| [rewrite](https://github.com/openrewrite/rewrite) | 3.5k | Refactoring | AST | +| [codebase-memory-mcp](https://github.com/DeusData/codebase-memory-mcp) | 2.7k | Read | AST + SQLite | +| [cocoindex-code](https://github.com/cocoindex-io/cocoindex-code) | 1.7k | Read | Embedding | +| [grepai](https://github.com/yoanbernabeu/grepai) | 1.7k | Read | BM25 / Grep | +| [CodeGraphContext](https://github.com/CodeGraphContext/CodeGraphContext) | 3.7k | Read | Graph DB | + + +## Resources + +### [ManoMano & Serena — Project Aegis](https://medium.com/manomano-tech/project-aegis-benchmarking-ai-agents-and-why-serena-is-our-new-must-have-311673db35dd) + +> Good size project for the test — information retrieval and refactoring. +> Slight cost increase, time decrease, but most importantly much better quality. +> Used fewer subagents (subagents are long and bad because of context reset). +> Claude Code LSP was bad. CC is still better for quick exploration. +> Refactoring is where Serena shines. + +--- + +### [Codebase-Memory: Tree-Sitter-Based Knowledge Graphs for LLM Code Exploration via MCP](https://arxiv.org/abs/2603.27277v1) +*28 Mar 2026 — 2 citations* + +> 10% less accuracy for 10× fewer tokens and 100× faster. +> Uses TreeSitter + SQLite. Doesn't require specialized models, supports many +> languages, auto syncs. The questions in their benchmarks are very precise +> (naming functions) — my use of the tool is much less directed. +> They disregard Code Property Graphs and CodeQL as unfit for LLM (Claude says +> CodeQL needs a compilation step; others need Neo4j and upkeep). I wonder why +> — that seems doable to me. TreeSitter is incremental, so obviously easier. +> They explicitly compare to an "Explorer Agent" and to other groups (RAG, +> Repo-Map, Graph+LLM). Good at exploring relations. The core opposition: +> embedding retrieval (TEXT) vs structural retrieval (AST). + +--- + +### [Code Intelligence Tools — rywalker.com](https://rywalker.com/research/code-intelligence-tools) + +> A useful tiering: +> - Knowledge graph engines — GitNexus, CodeGraphContext +> - MCP code search — Octocode, CodePathFinder +> - Context packing — Repomix, code2prompt +> +> It's true that I quite often experience broken tests after refactoring. + +--- + +### [Retrieval-Augmented Code Generation: A Survey with Focus on Repository-Level Approaches](https://arxiv.org/abs/2510.04905) +*6 Oct 2025 — 19 citations* + +> Real-life work is more complex than coding challenges — Repository-Level Code +> Generation. They review context construction, retrieval optimization, generation, +> and environment interaction. +> Graph vs non-graph approaches. Non-graph: as text, similarity (BM25, Jaccard). +> One-shot retrieval vs dynamic agentic exploration. Some systems use RL-trained +> models to improve retrieval. +> Increasing model context size provides similar benefits as RAG. +> GRAPH are the next in town vs TEXT — that doesn't mean it's better. + +--- + +### [Subagents — Simon Willison](https://simonwillison.net/guides/agentic-engineering-patterns/subagents/) + +> A way to work around limited context. Even 1M context is worse than 200k context +> used well. Claude Code Explore subagent avoids many tokens in the parent context. +> With current caching, it feels like it doesn't cost more? + +--- + +### [SWE-agent: Agent-Computer Interfaces Enable Automated Software Engineering](https://arxiv.org/abs/2405.15793) +*6 May 2024 — 1159 citations* + +> Now an old paper. They introduced the need for tools (shell was too hard), +> simple compact commands, concise feedback, guardrails (read before edit…). +> Things that have aged: edit doesn't have read-before, partial file edit shows +> lint errors. +> They beat the RAG-based previous record — but it costs 10× more. + +--- + +### [code-review-graph](https://github.com/tirth8205/code-review-graph) + +> Claim around 50× saved tokens per analysis request. +> I don't know if this works well for editing. + +--- + +### [grepai — Benchmark vs grep / Claude Code](https://yoanbernabeu.github.io/grepai/blog/benchmark-grepai-vs-grep-claude-code/) + +> Here too, 50×. Note — it has only a −25% impact on cost. +> Input tokens are not so expensive. + + +{{< alert info "LLM usage in this article" >}} +- Converting bullet lists to markdown tables +- Structuring research notes into blockquotes +- Replacing footnotes with cross-links to resource analyses +- Correcting spelling and grammar throughout +- Acting as Publisher +{{< /alert >}} |