We fix the agent's skill issueit reads too much code to find too little
A small lab building the context layer for AI coding agents: semantic code search, call graphs and MCP servers. We publish what we build and how we measure it.
Products
What the lab shipspick one to see how it works
Each product has its own page with the details, numbers and pricing.
Finder
A context engine for your repo. Semantic code search over MCP that answers an agent with exact snippets and file:line citations instead of whole files.
Finder Kit
Ships with Finder. Skills that teach an agent how to use the index at each stage of a task: explore, frame, plan, implement, test, review.
Technology
Search by meaning finds the starta call graph finds everything connected to it
The rest is making both fast, bounded and honest about what they know.
Semantic retrieval
Embeddings plus a reranker match a plain-words question to code that never uses those words.
Call graphs
Tree-sitter parsers extract symbols and call edges for 21 languages, so a hit expands to callers, callees and tests.
MCP servers
One tool per kind of question, each returning bounded, read-only output. Works with any MCP client.
Context packs
Compact answers where every item is graded: proven from source, inferred from the graph, or marked as a gap.
Impact analysis
Before an edit, the graph reports what a change to a symbol breaks: affected callers, files and tests.
Local index
The index and call graph live on your machine, per workspace, and only for folders you approve.
How we work
Three ruleswe hold ourselves to
Measure it
Claims come with a benchmark, the method behind it, and the caveats that would make you trust it less.
Cite it
Every answer points at a file and a line. If the tool does not know, it says so instead of guessing.
Spend less
Tokens are the budget. A tool earns its place only if the agent reads less code to get the same answer.
Writing
Notes from the labwhat we learned building it
What is a context engine?
GuideSemantic code search vs grep for AI coding agents
GuideImpact analysis for AI agents: know the blast radius before you edit
GuideHow to cut token usage in Claude Code, Cursor and Codex
FAQ
Questions people askbefore they try it
What is Skillissuelab?
A small lab that builds developer tools for AI coding agents. We focus on one problem: agents read too much code to find too little. Finder is our first product.
Which products can I use today?
Finder, a semantic code search server that speaks MCP, and Finder Kit, the workflow skills that ship with it.
Is Finder free to try?
Yes. Every account starts with 40 free credits, enough to index a project and explore it. See the pricing page for the paid packs.
How does payment work?
Finder is paid for with one-time credit packs. There is no subscription and nothing renews automatically. Prices are listed on the pricing page. Checkout is handled by a third-party payment provider; we never receive your full card details.
Does my code leave my machine?
The index and call graph are built and stored on your machine, only for folders you approve. To search by meaning, Finder sends code snippets and your questions through our gateway to an embedding provider. The gateway does not store those request bodies, and we do not train on your code. The privacy policy has the details.
Can I get a refund?
Yes. Ask within 14 days of purchase and we refund a pack whose credits have not been used. See the refund policy.
How do I contact you?
Email [email protected]. We reply within 3 business days.