IRIS
Code Intelligence for AI Agents
Stop copy-pasting documentation. Give your coding agents instant access to any codebase's knowledge.
Automatic Cloning & Indexing
Point to any public or private GitHub repo. IRIS clones it, parses the structure, and indexes every source file for semantic search.
Semantic Code Search
Ask questions in natural language. Find implementations, understand patterns, trace dependencies — powered by vector embeddings.
Coding Agent Integration
Works with Claude Code, OpenCode, Cursor, and any agent supporting IRIS protocols. Query repos mid-task without leaving your workflow.
Local Vector DB (SQLite)
All embeddings stored locally in SQLite with Ollama for generation. No cloud dependencies — fully offline-capable and private.
Multi-Repo Knowledge Graph
Reference multiple repos in one query. Cross-repo dependency analysis, pattern matching, and unified semantic search.
Auto-Sync on Push
GitHub webhooks trigger re-indexing when repos update. Your knowledge base stays current with every commit.
How It Works
From GitHub URL to queryable knowledge base in under a minute.
Declare Repo References
Add GitHub URLs to your agent config or create a standalone repo agent. Supports public HTTPS and private SSH repos.
Automatic Indexing
IRIS clones the repo, parses 18+ file types, chunks code by function and class, and generates 768-dim vector embeddings via Ollama.
Query from Coding Agents
Claude Code, OpenCode, or Cursor can now ask questions about the repo. Get answers with file paths, line numbers, and scored matches.
Bridge to Sub-Agents
Your coding agent can delegate to repo-specific sub-agents for deeper analysis, cross-repo search, and context gathering.
How Developers Use Repos as Agents
Every codebase you bring in becomes a knowledgeable collaborator.
Onboarding Acceleration
New developers ask questions to the codebase instead of senior engineers. "How is authentication implemented?" "Where are the payment webhooks handled?"
Coding Agent Context
Claude Code references your team's internal libraries, design patterns, and API clients while building new features — no copy-pasting docs.
Dependency Analysis
Query how a third-party library is used across your monorepo. Find all usages of a deprecated API, trace data flows, map function call chains.
Documentation as Code
The codebase IS the documentation. Agents read the actual implementation — no more out-of-sync docs, stale READMEs, or incomplete wikis.
Security & Privacy
Your source code is your most valuable IP. Repos as Agents is built with security as a first principle.
Files Never Executed
Source files are read-only — chunked and embedded as text. No code is ever evaluated, compiled, or run during indexing.
All Data Stays Local
Zero cloud uploads. Embeddings generated locally via Ollama. Vectors stored in SQLite on your filesystem. No telemetry, no phone-home.
Auth-Protected Endpoints
All bridge API endpoints require X-Bridge-Key authentication. Disk usage capped at 10 repos. Max 1GB vector storage with configurable limits.
Turn Your Repos into Knowledge Bases
Every codebase you bring in becomes a knowledgeable collaborator. Clone, index, query — all locally, all private.