
Syke
io.github.khalomskyPublished by GitHub account khalomskyv1.5.4Updated Oct 24, 2026
AI code impact analysis — dependency graphs, cascade detection, and build gates for AI agents
Overview
SYKE is a local MCP server that maps a codebase's dependency graph and lets an AI agent check whether its edits break dependents before building or deploying.
- What you can do
- Check build safety: run a pre-build gate that returns PASS/WARN/FAIL, or a quick HIGH/MEDIUM/LOW/NONE risk verdict.
- Inspect dependencies: list a file's internal imports and analyze the full impact of a change, including circular clusters and risk scores.
- Rank and refresh the graph: find the most important hub files by PageRank and rebuild the dependency graph after changes.
- AI and monitoring: run semantic analysis of a file and its dependents through a chosen AI provider, and get alerts on changes that may break dependents.
- Visualize: a local web dashboard shows an interactive dependency graph with risk indicators.
- When to use it
- Useful when an AI coding agent edits a multi-file project and you want an automated check that nothing downstream breaks before committing or deploying. Best for teams that want dependency impact analysis and a build gate wired into their agent workflow.
- Requirements
- Runs locally via npx from the npm package @syke1/mcp-server, so Node.js is needed; desktop only. Optional config in ~/.syke/config.json or environment variables: SYKE_LICENSE_KEY for Pro features, and one AI provider key (GEMINI_KEY, OPENAI_KEY or ANTHROPIC_KEY) for semantic analysis. The dashboard listens on localhost port 3333 by default.
Installation
In SourceWeft
- Open Syke in the dashboard and add it to a workspace.
- Enable the server for the chats that should use its tools.
Desktop only via STDIO. STDIO servers start a local process, so they need the SourceWeft desktop host.
Other MCP clients
Follow the launch instructions in the repository.
README
SYKE — Your Codebase Has a Pulse
AI code impact analysis MCP server. SYKE monitors every file your AI touches, maps dependency graphs, detects cascading breakage, and gates builds before damage spreads.
Works with Claude Code, Cursor, Windsurf, and any MCP-compatible AI coding agent.
How It Works
- On startup, SYKE scans your source directory and builds a complete dependency graph using static import analysis.
- Your AI agent modifies files freely — no interruptions during normal work.
- Before build/deploy, the AI calls
gate_buildto check if all changes are safe. - If dependencies break, SYKE detects cascading failures and blocks the build with a
FAILverdict. - The dashboard shows a real-time visualization of your dependency graph with risk indicators.
SYKE is a safety net, not a gatekeeper. It doesn't block your AI while working — it catches what your AI missed before you ship.
Quick Start
1. Create config file
Create ~/.syke/config.json:
Get your license key at syke.cloud/dashboard. You only need ONE AI key. Supported:
geminiKey,openaiKey,anthropicKey.
2. Register MCP server
Claude Code:
Cursor (.cursor/mcp.json):
Windsurf (~/.codeium/windsurf/mcp_config.json):
Windows note: If
npxis not found, use the full path:"command": "C:\\Program Files\\nodejs\\npx.cmd"
3. Add build gate to your project
Add this line to your project's CLAUDE.md (or equivalent AI instruction file):
This ensures your AI agent automatically runs SYKE's safety check after every task — no manual prompting needed.
4. Restart your AI agent
SYKE auto-detects your project language and builds the dependency graph on startup. Open http://localhost:3333 to see your live dashboard.
Features
8 MCP Tools
Multi-AI Provider Support
SYKE supports three AI providers for semantic analysis. Bring your own key:
Auto-selection: SYKE uses the first available key (Gemini > OpenAI > Anthropic).
Force provider: Set aiProvider in config (or SYKE_AI_PROVIDER env var) to override.
Advanced Graph Algorithms
SYKE goes beyond simple dependency counting. Five production-grade algorithms work together to deliver precise, fast, and context-rich impact analysis — all running locally with zero AI token cost.
1. SCC Condensation + Topological Sort
Circular dependencies are the #1 source of misleading impact analysis. SYKE uses Tarjan's algorithm to detect all Strongly Connected Components, condenses them into a clean DAG, then runs topological sort to compute correct cascade levels.
- O(V+E) computation — runs in single-digit milliseconds
- Every SCC with size > 1 is flagged as a circular dependency cluster
- Cascade levels are accurate even in heavily cyclic codebases
2. Composite Risk Scoring
Five signals combined into a single 0–1 risk score:
AI agents can now make threshold decisions: proceed if < 0.3, warn if 0.3–0.7, block if > 0.7.
3. Historical Change Coupling
Static imports miss hidden dependencies — files that always change together but have no import relationship. SYKE mines your git history (last 500 commits) to find these logical couplings.
- Catches 15–30% of impacted files that static analysis misses entirely
- Filters mega-commits (>20 files) to avoid noise
- 5-minute cache with auto-refresh
4. PageRank for File Importance
Simple fan-in counts treat all dependents equally. PageRank computes recursive importance — a file imported by many important files ranks higher than one imported by many leaf files.
- Standard Power Iteration with damping factor 0.85
- Precomputed at startup, incrementally updated on file changes
- Every file gets a rank position and percentile (e.g., "rank #3 of 245, 99th percentile")
5. Incremental Graph Updates + Memoized Queries
For large codebases (10K+ files), full graph rebuilds are too slow. SYKE now updates only the changed file's edges and invalidates only the affected cache entries.
- Reverse index enables O(affected) cache invalidation instead of O(cache_size)
- SCC and PageRank recompute after edge changes (still < 100ms for 10K files)
- 500-entry LRU cache with hit/miss diagnostics
Language Support
Auto-detected, zero-config: Dart/Flutter, TypeScript/JavaScript, Python, Go, Rust, Java, C++, Ruby.
Web Dashboard
Live dependency graph visualization at localhost:3333 with:
- Interactive 3D node graph (click any file to see its connections)
- Real-time cascade monitoring
- Risk level indicators
- Server offline detection with auto-reconnect
Configuration
SYKE reads from ~/.syke/config.json (primary) with environment variable overrides:
Full config example (~/.syke/config.json):
Recommended Workflow
Founding 100 — Free Pro for Early Adopters
We're giving the first 100 developers full Pro access for 30 days — no credit card, no strings.
What you get:
- All 8 MCP tools with advanced algorithms (SCC, PageRank, Risk Scoring, Git Coupling)
- Unlimited files, multi-project support
- Real-time cascade monitoring + web dashboard
- AI semantic analysis (BYOK — Gemini, OpenAI, or Claude)
How to claim:
- Sign up at syke.cloud
- Star this repo
- Click "I Starred" in your dashboard → 30 days Pro unlocked
Spots are limited. Once they're gone, they're gone.
Source CodeThis repository contains the Free tier source code — the core dependency graph engine, language plugins, and 3 free MCP tools (gate_build, check_safe, get_dependencies).Pro and Cortex features (advanced algorithms, AI analysis, real-time monitoring, web dashboard) are included in the npm package as compiled code.
License
You can use, modify, and distribute SYKE freely, with two limitations:
- No managed service — You cannot offer SYKE as a hosted/managed service to third parties.
- No license key circumvention — You cannot remove, disable, or bypass the license key functionality.
Source: README.md at commit d68bef1
Tools
0Version history
1- v1.5.4LatestSep 16, 2026
Score
40% of the applicable checks met
How complete this listing is, computed from the registry entry, the endpoint and public sources. Not a rating of the server and not a statement that it is safe or working today: reachability is a dated check, and no installs or reviews go into the score.
- Reached in the last 30 days25 pts
Keep the endpoint answering the MCP handshake; SourceWeft probes it every 30 days.
- Auth declared10 pts
Declare the credential headers in the registry entry, or keep the endpoint answering the probe so its auth can be classified.
- Tools listed15 pts
Answer tools/list without a credential, or ship a package that starts in a sandbox without one.
- DescriptionRequired15 pts
- Logo10 pts
- Documentation15 pts
- License10 pts
Declare a license in the repository or the registry entry.