Oct 1, 2026
C# code graphs: what CGraph extracts, and what it misses
CGraph builds a C# code graph from classes, methods, and calls, but properties, constructors, and interfaces leave real gaps. Measured on MediatR and Polly.
→
— 34 posts tagged
code intelligence
Oct 1, 2026
TypeScript barrel files: following export * in a code graph
An import through an index.ts barrel used to stop at the barrel, so impact analysis never reached real users. How CGraph follows export * to the declaring file.
→
Sep 4, 2026
Why a code graph can't resolve obj.method(): receiver types
A call graph resolves obj.method() only when something proves what obj is. The five tiers of evidence, and why Java call sites supply it while Rust ones do not.
→
Aug 27, 2026
The benchmark oracle that scored every miss as noise
git show --name-only prints nothing for a merge commit. That one fact turned our dependency oracle into a machine that cleared every miss on zero evidence.
→
Aug 27, 2026
Dogfooding test impact: Blastline on CGraph's own PRs
CGraph builds a graph of itself on every PR and lets Blastline pick the tests. Eleven PRs in: six real subsets, five "run everything" — all five from one file.
→
Aug 26, 2026
When a code graph should refuse to guess a call target
A wrong CALLS edge is a false dependent forever. CGraph resolves a call only when one candidate is provable, drops the rest, and counts every drop it makes.
→
Aug 25, 2026
Java interface dispatch: resolving s.area() in a code graph
Shape s = ...; s.area() produced no edge in CGraph, because the name was ambiguous. How interface dispatch resolves it, the overload regression it caused, and what it costs.
→
Aug 25, 2026
Kotlin code graphs: when 'supported' extracts nothing
CGraph listed Kotlin as a supported language while extracting zero symbols from it. The cause was a tree-sitter grammar with no named fields — and what finally measured it.
→
Aug 24, 2026
Your coding agent says it updated every caller. Check.
Your AI agent claims it updated every call site. blastline check queries a code graph for what still references the symbol, with file:line evidence — and refuses to certify.
→
Aug 20, 2026
Blastline v0.4–v0.8: a standing safety audit, and the barrier that fixed Go
Four releases since launch: cross-repo seams, C/C++, base-graph automation, a safety-audit loop that runs on every merge, and the dispatch barrier that cut Go selections by a third without losing a single test.
→
Aug 20, 2026
Our test-impact benchmark was lying — so we built an oracle that could not
Blastline scored 6/12 on tokio and it looked like a recall crisis. A dependency-filtered oracle showed five of the six "misses" depended on nothing — and chasing the one real miss uncovered a macro-blindness bug in CGraph. Tokio reachability went 0.59 → 0.91.
→
Aug 19, 2026
Blastline: graph-backed test impact analysis for CI
Blastline picks the tests a diff actually reaches by querying a CGraph code graph — deterministic, fail-open, benchmarked on 80 historical commits. Now open source.
→
Aug 18, 2026
Rust code graphs: CGraph adds its 13th language
CGraph now extracts Rust — structs, traits, functions, and call edges including method calls — via a pinned tree-sitter grammar, with honest v1 limits.
→
Aug 14, 2026
Stale index? How a code graph proves its answers are fresh
Every CGraph answer is stamped with a merkle content root of the indexed source, and reads can pin that root — so a stale index errors instead of lying.
→
Aug 12, 2026
Check a refactor before you make it: query, impact, path
A safe-refactoring workflow on a real repo: pin the node with query, size the blast radius with impact, explain surprise dependents with path — before editing.
→
Aug 12, 2026
Visualize a codebase graph on the web: one HTML file
How our live demo embeds an interactive graph of 1,300 nodes: a self-contained graph.html export, a click-to-load sandboxed iframe, and a provenance gate.
→
Aug 10, 2026
Fold-in vs full rescan: incremental code graph indexing
How CGraph keeps a code graph current as you edit: content-hashed fold-ins within a couple of seconds, and why a branch switch becomes one full rescan.
→
Aug 4, 2026
One daemon per repo, always warm: cgraph daemon install
cgraph daemon install keeps a resident code-graph daemon warm for every repo you track — auto-discovered, restarted on crash, started at login via launchd.
→
Jul 28, 2026
Measuring AI agent tool usage: the op-stats ledger
Which MCP tools does your coding agent actually use? CGraph records every daemon op in a ledger; cgraph stats rolls up counts, latency, and zero-hit rate.
→
Jul 18, 2026
A code graph as agent memory across sessions
Checkpoint a coding session into the code graph and recall it after /clear or a daemon restart. How CGraph stores durable, code-linked memory that never pollutes retrieval.
→
Jul 18, 2026
Enriching a code graph with LLM-authored knowledge
Layer prose, concepts, and design intent onto a deterministic code graph — without the binary ever calling an LLM. How CGraph plans chunks and merges host-authored semantic fragments.
→
Jul 16, 2026
Cross-service impact analysis with cgraph seam
Fuse per-repo code graphs into one queryable cross-service graph: declare the wire contract, anchor it to real code, and trace blast radius across services.
→
Jul 15, 2026
Hardening a local daemon: four bugs that passed every test
A localhost socket is still a trust boundary. How four latent bugs in CGraph's resident daemon survived a green test suite, and what fixing them looked like.
→
Jul 12, 2026
Budgeted context: just enough of your repo for an LLM
RAG for code: how CGraph gathers a neighborhood around a symbol and packs it into an LLM token budget with a knapsack, returning an inspectable result.
→
Jul 12, 2026
Code graphs vs LSP vs ctags: what each is good for
Code graphs, LSP, and ctags answer different questions about a codebase. A decision guide to when each one wins, and where the three usefully overlap.
→
Jul 12, 2026
Deterministic code-graph diffing on every pull request
CGraph extraction is deterministic: the same tree yields the same graph. That makes a code graph diffable in CI with the one-shot CLI and plain diff tooling.
→
Jul 12, 2026
Exporting your codebase to Obsidian, Cypher, or SVG
CGraph writes the same code graph in six shapes — JSON, HTML, SVG, an Obsidian vault, Cypher, and a call-flow view. Here is when to reach for each export.
→
Jul 12, 2026
Feeding your codebase to Claude Code: graphs vs file dumps
Give a coding agent your repo as a queryable graph over MCP instead of dumping raw files — precise answers, fewer tokens, and a one-line Claude Code setup.
→
Jul 12, 2026
Impact analysis in ~10ms: everything a change touches
Find the blast radius of a code change in one query. CGraph keeps a resolved graph warm in a daemon, so "what breaks if I change this?" returns in about 10ms.
→
Jul 12, 2026
The MCP tools CGraph gives your coding agent
CGraph exposes eight MCP tools that let a coding agent query a codebase as a graph. What each tool answers, and when an agent should reach for each one.
→
Jul 12, 2026
How CGraph extracts a code graph across 11 languages
CGraph parses 11 languages with vendored tree-sitter grammars, then resolves imports, calls, and relations into one deterministic code graph. Stage by stage.
→
Jul 10, 2026
Benchmarking CGraph: native code graphs vs Graphify and grep
We measured CGraph against Graphify and against the grep-and-read loop agents use today: 12–198x faster builds, a 15.8x faster warm query, and ~82% fewer tokens on realistic navigation tasks — with the caveats that matter.
→
Feb 7, 2025
Contributing to CGraph: A Guide to Making an Impact
A practical guide to contributing to CGraph — pick a good first issue, build it locally with CMake and vcpkg, keep extraction deterministic, and open a PR we can review quickly.
→
Feb 7, 2025
Open-Source Code Intelligence: A CGraph Alternative to Proprietary Tools
CGraph is an open-source code-intelligence engine that builds a queryable graph of any repository — a self-hosted alternative to proprietary code-search and analysis tools.
→