DiffGate

A deterministic guardrail your coding agent runs on itself — before the code reaches your disk.

Every other check fires too late. A review bot needs a PR. A pre-commit hook runs after the agent has finished and moved on. DiffGate is an agent hook: your agent calls it over MCP, gets back structured findings in milliseconds for zero LLM tokens, and fixes the problem while it still has the context — then the same engine, same verdict runs in your editor, your pre-commit hook, and CI. Not a model grading its own homework: the same input always produces the same answer.

It stays quiet by construction. It grades only the lines that changed (🟢 merge · 🟡 glance · 🟠 verify), runs your tests only when a change earns it, and blocks only when it's earned — 0 false blocks on a public, versioned corpus you can rerun yourself with diffgate bench (BENCHMARK.md). Everyone claims low noise; this one ships the corpus.

Why an agent hook, not another review bot

Every check in the pipeline fires after the code exists, and each one is later than the last:

Being early is not the only thing that matters, though — it has to be quiet, or the agent learns to ignore it. Review bots comment after the PR exists; linters and scanners flag everything they see. Neither guarantees the risky line gets discussed: we scanned 350 merged AI-assisted PRs — of the 109 with flagged AI-attributed changes, only 3 drew public discussion from any human besides the author. DiffGate decides what deserves your attention, your tests, or a block, and stays quiet otherwise. That's the whole product:

• Diff-scoped. Findings report only on the lines that changed, against the committed baseline — no whole-file noise, no re-litigating code you didn't touch.
• Tiered triage, not a flat list. Three tiers route attention: green merges, yellow is a glance, orange is gated.
• The gate runs your tests — selectively. On an orange change, DiffGate runs your testCommand and shows the real exit code and output. Green and yellow pass instantly. The pre-commit hook is fast because tests fire only when a change is genuinely high-impact.
• Earns the right to block. Broad cross-language injection findings stay advisory on their own; they escalate to a blocking finding only when the optional code graph proves the sink is reachable from an untrusted entry point (an HTTP/event handler). Recall from the rules, the right to block from the graph.
• Change-impact aware. With an optional code graph, a finding carries its cross-file blast radius — caller counts, suggested reviewers, untested call sites — and an exported symbol nobody calls is de-escalated. Cross-file context makes reviews quieter, not louder.
• Fast. A review runs in milliseconds on the changed lines — quick enough to sit in the agent and editor inner loop, not only in CI.
• Provably low-noise. diffgate bench runs a versioned corpus offline: 100% precision / 0 false blocks on clean changes. Reproduce it yourself — that's the point of shipping the corpus. See BENCHMARK.md.

The measurement is reproducible

The 0% / 13% numbers aren't a marketing line — they come from a scripted experiment (four models from local to frontier, greenfield vs. edit mode, Wilson confidence intervals) that you can rerun with diffgate marginal. Methodology, per-model tables, and caveats: docs/MEASUREMENT.md. DiffGate's security rules are tuned to that measured residue, not to maximizing rule count.

Quick start

npm install -g diffgate-review
cd your-repo
diffgate init                    # auto-detects language + test command, writes .diffgate.json
diffgate check --since=HEAD~20   # see what it catches in your own history — no PR required
diffgate check                   # review your pending changes right now

No git history or uncommitted changes yet? See the output on bundled examples first:

diffgate init --demo   # live scan, no config or git changes needed

The surfaces (one shared engine, one verdict)

1. In your coding agent (via MCP)

The highest-leverage spot: the agent self-checks generated code before it's written to disk, gets back structured findings (zero LLM tokens), and surfaces what it corrected (original + fix + why) instead of silently rewriting. A trustworthy, deterministic self-check is what makes it safe to grant the agent more autonomy.

# Claude Code — one command:
claude mcp add diffgate -- diffgate mcp

# One-click via Smithery (zero config):
npx @smithery/cli install diffgate-review --client claude

# Cursor — add to MCP settings:
# { "diffgate": { "command": "diffgate", "args": ["mcp"] } }

Or one-click in Claude Desktop: download diffgate.mcpb and open it. The server also exposes prompts and resources; see MCP.md.

2. In your editor (VS Code / Cursor)

Inline squiggles on changed lines, hover cards (why · who owns it · quick-fix), a Risk Review tree, a status-bar summary, and Deep Review (agentic blast-radius analysis for orange findings). The same verdict you'd get from the CLI, on the diff you're reviewing.

Install from the VS Code Marketplace or Open VSX (Cursor / Windsurf / Gitpod).

3. On the command line — and in CI

diffgate check reviews your diff and exits non-zero on high-impact findings: a pre-commit hook locally, the same gate in your pipeline.

diffgate install-hook  # adds .git/hooks/pre-commit; only runs tests on 🟠 orange changes

The local loop is the wedge — fix while the context is fresh — and the same engine runs as a PR gate so the verdict carries to where it's enforced for the whole team. See docs/TEAM.md for the GitHub Action, shared learnings, and org policy packs. CI runs can optionally layer an external scanner (Semgrep) through the same gate for broader language coverage — advisory-only, off by default (docs/CONFIG.md).

Common commands:

diffgate check                 # review pending changes (the gate)
diffgate check --staged        # staged-only (pre-commit)
diffgate check --since=HEAD~20 # audit recent history, per-commit (see below)
diffgate check --agent         # machine verdict for coding agents
diffgate scan <path>           # analyze files directly (no git needed)
diffgate watch                 # live review as you edit
diffgate guidelines            # review diff against AGENTS.md / CLAUDE.md etc.
diffgate feedback <rule> <f> <l> --dismiss   # suppress a false positive (shared via git)
diffgate mcp                   # start the MCP stdio server

Audit recent AI-authored history. Point check at commits already in your log — each finding is attributed to a specific commit, so you get a story, not a repo-wide report card:

diffgate check --since=HEAD~20        # last 20 commits, one block per commit
diffgate check --since="2 weeks ago"  # by date instead of a rev
diffgate check --ai-authored          # only agent commits (Claude/Copilot/Cursor/… — heuristic)
diffgate check --author="Claude"      # matches author *and* Co-authored-by trailers
diffgate check <sha>                  # a single commit by hash

History mode is report-only (it audits the past — it never runs your test command or blocks a commit) and honors --json and --limit=<n> (default 50). Merge commits are skipped.

Run diffgate --help for the full list (report, bench, stats, graph, marginal, …).

How it works

• Diff-aware: git diff (CLI) or an in-memory LCS diff (editor, accurate on unsaved buffers) finds changed lines; findings only report on those lines.
• Real AST where it counts: @babel/parser (JS/TS) and tree-sitter (Python, PHP, Go, Ruby, Java, C#, Kotlin — via WASM, no native build) power precise rules: deprecated calls aren't matched inside comments or strings, exported-signature changes are detected structurally, and SQL injection is sink-targeted, parameter-aware, and sanitizer-aware — cur.execute(f"… {uid}") / $pdo->query("… $id") block, while cur.execute("… %s", (uid,)), $pdo->prepare("… ?"), a single-quoted '… $id', and a SELECT in a log line don't.
• A deterministic floor everywhere else: comment-aware pattern rules for secrets, destructive/schema changes, auth/crypto, dynamic execution / shell-out, raw queries, and network calls across Go, Java, Ruby, and any text. Commented-out code (# os.system(x)) isn't flagged; a secret committed inside a comment still is. Docs/prose files (.md, .rst, …) are held to the same standard: the word "oauth2-provider" in a changelog isn't auth code, but a key pasted in a README is still a leak.
• Earned blocking: broad cross-language injection advisories for the non-AST languages (Ruby #{}, Go/Ruby shell-out) escalate to blocking only when the optional code graph proves reachability from an untrusted entry point — community CodeGraph, no Pro taint engine required. (JS/TS, Python, and PHP block on local AST evidence and don't need this.)
• The gate: on a high-impact change, DiffGate runs your testCommand and shows the actual exit code and output.
• Learnings: diffgate feedback records dismiss/confirm verdicts; dismissed findings (same rule + same code) are suppressed everywhere. Stored in .diffgate/learnings.json; commit it to share across the team.
• Optional add-ons: a provider-agnostic AI layer (plain-English explanations + fixes) and a cross-file blast-radius pass via an optional code graph. Both are off by default and degrade gracefully to a no-op.

Engine layout: src/core (shared) · src/cli.ts (CLI) · src/mcp.ts (MCP) · extension/ (VS Code).

Coverage scales with language

How deeply DiffGate analyzes a change depends on the file's language — be explicit about this so you can calibrate how much to trust a clean result.

Sink classes per Deep-AST language (full detail — every sanitizer and safe-form, plus the code-graph boundary — in docs/SCOPE.md):

• Python (7) — SQL · XSS · path traversal · CORS · command · code · deserialization
• PHP (8) — SQL · command · code · file inclusion · deserialization · XSS · path traversal · CORS
• Go (4) — SQL · command · path traversal · CORS
• Ruby (6) — SQL · command · code · deserialization · XSS · CORS
• Java (6) — SQL · command · deserialization · path traversal · XXE · CORS
• C# (7) — SQL · command · deserialization · path traversal · XSS · XXE · CORS
• Kotlin (6) — SQL · command · deserialization · path traversal · XXE · CORS

SSRF is a cross-language advisory across all eight Deep-AST languages (a request-tainted URL into an outbound-request sink; library-qualified and tainted-only, so static/config URLs aren't flagged). XXE covers the JVM (Java, Kotlin) and .NET (C#), suppressed when the file shows recognized hardening. Permissive CORS now also covers all eight — wildcard Access-Control-Allow-Origin, allow-all framework configs (gin/rs-cors, Spring @CrossOrigin, ASP.NET AllowAnyOrigin(), Ktor anyHost(), rack-cors), and request-reflected origins; explicit allowlists aren't flagged.

Fast by design — and scoped to match. A review runs in milliseconds on the changed lines, which is exactly what lets the same check sit in the agent and editor inner loop. That speed is a deliberate trade: DiffGate is the deterministic gate on the diff, not an exhaustive whole-repo taint engine. Coverage is per-language (deep where there's an AST, a pattern floor elsewhere), the security rules are tuned to the residue agents actually ship rather than to maximize raw rule count, and a clean result means "nothing matched at this language's tier," not "proven safe." For deep cross-file taint analysis across many languages, pair it with a dedicated SAST. Full per-language detail and the code-graph boundary: docs/SCOPE.md.

Configuration

diffgate init writes a tailored .diffgate.json at your repo root. Minimal example:

{
  "testCommand": "npm test",          // run for orange changes (the gate)
  "gate": { "mode": "working", "failOn": "orange" },
  "deprecated": [
    { "pattern": "StripeClient.charge", "replacedBy": "StripeClient.createPaymentIntent" }
  ]
}

Any rule — built-in or custom — can be path-scoped with include/exclude globs, the escape hatch for the one file where a forbidden idiom is legitimate (e.g. process.env inside the config loader itself). Full schema, the built-in rule table, LLM providers, and per-rule tuning: docs/CONFIG.md.

More

• The PR was reviewed. The risky line wasn't. — our PR-trail study: 350 merged AI-assisted PRs, 109 with flagged AI-attributed changes, 3 with public human discussion beyond the author.
• docs/SCOPE.md: per-language coverage tiers (deep AST vs. pattern vs. text-only) and what the code graph does and doesn't do.
• docs/CONFIG.md: full .diffgate.json schema, all built-in rules, LLM providers, native precision & test-scope behavior.
• docs/TEAM.md: rolling DiffGate out to a team (GitHub Action / PR gate, shared learnings, org-wide policy packs, SOC 2 evidence, metrics for leaders).
• docs/CODE-GRAPH.md: optional cross-file blast radius (caller counts, suggested reviewers, test gaps, reachability, taint analysis).
• docs/STRUCTURAL-RULES.md: the structural pack — over-engineering rather than vulnerabilities (needless abstractions, pass-through wrappers, complexity spikes), all non-blocking, with per-language thresholds.
• docs/MEASUREMENT.md: what agents actually ship unprompted and how to reproduce it (diffgate marginal).
• MCP.md: MCP tools, prompts, resources, and AI configuration.

Try it

diffgate scan mock_project

You'll see green findings (logging), yellow findings (a deprecated call), and orange findings (a DROP COLUMN migration, a public export).

Tests

npm test    # builds the extension, runs the full unit/integration suite + extension smoke test

Support the project

If DiffGate caught something for you — or you just like the idea of a deterministic gate for agent code — star the repo ⭐. It's the signal that tells other people this is worth trying.

• 🐛 Found a false block, or a sink it missed? Open an issue — a false block is a bug we treat as P0.
• 💡 Want a language or rule covered? File a feature request with the idiom you'd like caught.
• 🔒 Security report? Please disclose privately — see SECURITY.md.

Contributing & License

See CONTRIBUTING.md. Apache 2.0; see LICENSE.