Satori
Provides scalable vector storage for indexed code chunks, enabling efficient similarity search.
Provides local embedding generation for semantic code search.
Click on "Install Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type
@followed by the MCP server name and your instructions, e.g., "@Satorisearch codebase in /workspace/my-app for 'user authentication flow'"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
Satori
Agent-safe code retrieval for developers who use MCP coding agents on real repos.
Satori indexes a repo, keeps that index fresh, and gives agents a fixed six-tool surface for finding code, opening exact spans, checking callers/callees, and reading bounded evidence before an edit. It is read-only from MCP: source edits stay in your normal editor or agent host.
Why Developers Use It
Find behavior by intent, not just filenames or exact tokens.
Keep search focused on runtime code before pulling in docs or tests.
Open exact files, line ranges, and symbols instead of dumping broad context.
Trace nearby callers/callees when sidecar data is ready.
Get explicit
requires_reindex, stale-state, and noise guidance instead of silent bad context.Install the MCP server and first-party workflow skills with one CLI command.
Related MCP server: navigation-agent-mcp
Packages
Package | Purpose |
| Indexing, AST chunking, embeddings, Milvus/Zilliz storage, retrieval, incremental sync |
| MCP server with the six agent-facing tools and lifecycle gates |
| Installer, doctor command, and shell access to MCP tools |
Quick Start
Install managed MCP config for your client:
npx -y @zokizuan/satori-cli@0.3.2 install --client codex
npx -y @zokizuan/satori-cli@0.3.2 install --client claude
npx -y @zokizuan/satori-cli@0.3.2 doctorThe installer writes Satori-managed config and copies the first-party skills:
satori-searchsatori-navigationsatori-indexing
Manual JSON-style MCP config:
{
"mcpServers": {
"satori": {
"command": "npx",
"args": ["-y", "@zokizuan/satori-mcp@4.10.1"],
"timeout": 180000,
"env": {
"EMBEDDING_PROVIDER": "VoyageAI",
"EMBEDDING_MODEL": "voyage-4-large",
"EMBEDDING_OUTPUT_DIMENSION": "1024",
"VOYAGEAI_API_KEY": "your-api-key",
"VOYAGEAI_RERANKER_MODEL": "rerank-2.5",
"MILVUS_ADDRESS": "your-milvus-endpoint",
"MILVUS_TOKEN": "your-milvus-token"
}
}
}
}Manual TOML-style config:
[mcp_servers.satori]
command = "npx"
args = ["-y", "@zokizuan/satori-mcp@4.10.1"]
startup_timeout_ms = 180000
env = { EMBEDDING_PROVIDER = "VoyageAI", EMBEDDING_MODEL = "voyage-4-large", EMBEDDING_OUTPUT_DIMENSION = "1024", VOYAGEAI_API_KEY = "your-api-key", VOYAGEAI_RERANKER_MODEL = "rerank-2.5", MILVUS_ADDRESS = "your-milvus-endpoint", MILVUS_TOKEN = "your-milvus-token" }Restart the MCP client after changing config.
Runtime Setup
Satori needs an embedding provider and a Milvus-compatible vector store before indexing.
Cloud quality start:
EMBEDDING_PROVIDER=VoyageAI
EMBEDDING_MODEL=voyage-4-large
EMBEDDING_OUTPUT_DIMENSION=1024
VOYAGEAI_API_KEY=your-api-key
VOYAGEAI_RERANKER_MODEL=rerank-2.5
MILVUS_ADDRESS=your-milvus-endpoint
MILVUS_TOKEN=your-milvus-tokenLocal-first start:
EMBEDDING_PROVIDER=Ollama
EMBEDDING_MODEL=nomic-embed-text
OLLAMA_HOST=http://127.0.0.1:11434
MILVUS_ADDRESS=localhost:19530Provider, model, dimension, vector store, and schema are part of the index fingerprint. If they change, Satori blocks search with requires_reindex until you rebuild the index.
Agent Workflow
Use Satori as the investigation layer before edits:
list_codebases
manage_index action="create" path="/absolute/path/to/repo"
search_codebase path="/absolute/path/to/repo" query="where is auth refresh handled"
file_outline path="/absolute/path/to/repo" file="src/auth.ts"
call_graph path="/absolute/path/to/repo" symbolRef={...} direction="both"
read_file path="/absolute/path/to/repo/src/auth.ts" start_line=1 end_line=160Default search behavior is developer-oriented:
scope="runtime"so docs/tests do not dominate first results.resultMode="grouped"andgroupBy="symbol"to reduce duplicate chunks.rankingMode="auto_changed_first"to prefer active work when safe.debug=falseunless you are inspecting ranking/filter behavior.
Six MCP Tools
Tool | Use it for |
| See indexed roots and their lifecycle buckets |
| Create, sync, reindex, inspect status, or explicitly clear indexes |
| Runtime-first semantic search with operators, grouping, freshness, and navigation hints |
| Read sidecar symbol outlines and resolve exact symbols without guessing |
| Traverse bounded caller/callee context from a search-provided |
| Read bounded files, ranges, annotations, or exact symbol spans |
What Satori Is Not
Not an agent framework.
Not a source-code write server.
Not a replacement for tests, typecheck, code review, or grep.
Not a promise that static call graph hints prove runtime or assertion coverage.
Satori gives the agent better evidence. It does not remove engineering judgment.
Repository Layout
packages/core indexing, retrieval, embeddings, vector store, sync
packages/mcp MCP server, tool schemas, lifecycle gates, generated tool docs
packages/cli managed installer, doctor, direct shell tool calls
docs/ behavior specs, feature inventory, architecture notes
satori-landing/ static website HTML sourceDevelopment
pnpm install
pnpm build
pnpm run versions:check
pnpm -C packages/mcp docs:check
pnpm -C packages/mcp manifest:check
pnpm --filter @zokizuan/satori-mcp test
pnpm --filter @zokizuan/satori-cli test
pnpm test:integrationMore Docs
License
MIT (c) Hamza (@ham-zax)
This server cannot be installed
Maintenance
Resources
Unclaimed servers have limited discoverability.
Looking for Admin?
If you are the server author, to access and configure the admin panel.
Latest Blog Posts
MCP directory API
We provide all the information about MCP servers via our MCP API.
curl -X GET 'https://glama.ai/api/mcp/v1/servers/ham-zax/satori'
If you have feedback or need assistance with the MCP directory API, please join our Discord server