Nacre MCP server

Nacre Your index. Your access rules. Your perimeter. Agents see exactly what they're allowed to see. nacre.work · Docs · Quickstart · Discussions

Nacre is a self-hosted knowledge index with fine-grained access control. Agents reach it over MCP, applications over a REST API. No chat interface, no company assistant — just the context layer underneath them.

Why

Vector search is a solved problem. What isn't solved: making sure an agent querying a company index sees exactly the documents the requesting user is cleared for — and being able to prove it to an auditor.

inherited top-down. write does not imply read; admin implies both. Document-level grants and deny rules are commercial — this build refuses them and says so, rather than accepting a rule it cannot propagate.

• Permissions that inherit. Workspaces → layers, with read/write/admin

HNSW traversal, not after ranking, so top_k returns k permitted results rather than k minus whatever got stripped out.

• Filtering happens inside the index. Access filters are applied during

and local STDIO for developer agents. Agents authenticate with a service account key; OAuth discovery is served (RFC 9728), and client registration is the authorization server's business rather than ours.

• MCP as a first-class surface. Streamable HTTP per the 2026-07-28 spec,

BM25 fused with reciprocal rank fusion, plus a cross-encoder rerank where a deployment configures one. Not offered as a differentiator — the line above is the honest one — but an agent searching for SQLSTATE 23505, an invoice number or a variable name needs the literal match, and a dense-only index does not reliably return it.

• Hybrid retrieval, because agents ask for identifiers. Dense vectors and

endpoint, bound per layer. Changing the model on an existing layer is a reindex that keeps search answering throughout, and it is gated on recall against a query set you supply before it switches over.

• Bring your own models. Embeddings through any OpenAI-compatible

walks a directory into it and searches — the four curl invocations the quickstart spells out, for when you want the result rather than the contract. A document that fails to index is a non-zero exit, so a nightly ingest cannot report success having indexed nothing.

• A command line client. npx @nacre.work/cli signs in, creates a layer,

• Stays inside your network. Docker Compose, no phone-home.

Quickstart

git clone https://github.com/nacre-work/nacre && cd nacre
cp .env.example .env
docker compose --profile minimal up -d

Full walkthrough: docs/quickstart.md.

On an Apple Silicon Mac, read docs/apple-silicon.md first — the images are arm64 and the stack is native, but the embedder is the one piece you run on the host.

Layout

packages/api        REST API and authorization service
packages/mcp        MCP server (Streamable HTTP + STDIO)
packages/worker     indexing pipeline: parse, chunk, embed
packages/core       data model, permission resolver, shared types
packages/sdk        TypeScript SDK
packages/admin      community admin UI
services/parser     Python sidecar: bytes → {text, blocks, metadata}
docs/               specifications — normative, and ahead of the code

State

Shortened. The full README is on GitHub.