linza-mcp 0.1.3

Local-first MCP sidecar for agent workspaces, artifacts, review cards, and context export.

LINZA — Local MCP Server for Agent Workspaces

It does not change your data. It changes how you see it.

LINZA solves one concrete problem: you have a folder of Markdown notes, and you want an agent to understand what is inside without risking accidental rewrites. The core idea is a review-gated sidecar: the human decides, the agent executes.

It is a local MCP server for notes, documents, articles, chats, logs, and drafts. LINZA reads a folder, builds a map of topics and relations, shows evidence-backed review cards, and stores its conclusions next to your files in .linza/linza.db. Your Markdown stays yours.

LINZA does not impose a fixed ontology. It does not rename notes by itself. It gives an agent a safer way to see structure, and gives you a calm review gate before anything is accepted.

Russian version

---

What It Is For

LINZA is useful when you already have material, but you do not yet have a safe way to let an agent reason about it.

• A Markdown note folder: Obsidian or any other .md directory.
• Incoming material: text, articles, chats, logs, JSON, DOCX, XLSX, PDF with an optional extractor.
• Research or writing workspaces: many themes, decisions, drafts, and traces of work.
• Agent workflows: context should survive between sessions, but the agent should not freely rewrite notes, skills, rules, or memory.

A good first run is intentionally small: connect the folder, index it, review 3-5 cards, accept a few good seed examples, then let LINZA grow in small preview batches.

index -> map -> review cards -> teach -> grow preview -> explicit apply

---

What You Will See

• Domains: meaning clusters LINZA detects in the workspace.
• Material types: notes, drafts, specs, cases, and recurring forms found from structure.
• Relations: what belongs together, what may cause or support something else.
• Patterns: recurring problems, terminology drift, topic gaps, possible contradictions.
• Memory: what future agents should recall, what might go stale, and what needs review.

Every serious card should answer the human question: why does LINZA think this? It carries evidence: notes, snippets, nearby chunks, relation labels, confidence, and write impact.

---

How It Differs

From Obsidian Graph View

Graph View shows links that already exist. LINZA tries to show what is not yet linked: hidden topics, possible relations, causal chains, recurring patterns, and review cards. It does not replace the Obsidian graph; it gives agents a working layer above a folder.

From Dataview And Auto-Tagging Plugins

Dataview is excellent when the structure is already written in YAML and links. LINZA proposes hypotheses first, shows evidence, and keeps acceptance behind a review gate. By default it is preview, not automation.

Installation

1. Install LINZA

python -m pip install linza-mcp

If you want LINZA to extract PDF text directly:

python -m pip install "linza-mcp[pdf]"

If you do not need PDF extraction, the normal install is enough. [pdf] adds the local pypdf extractor.

2. Choose a folder

LINZA works with any Markdown folder: an Obsidian vault, a project workspace, or a standalone document folder.

In the examples below, replace /absolute/path/to/workspace-or-vault with your own path.

3. Configure embeddings

For semantic search, LINZA needs a local embedding model.

The simplest local path is LM Studio:

1. Open LM Studio.
2. Download an embedding model, for example text-embedding-granite-embedding-278m-multilingual, nomic-embed-text-v1.5, or another embedding model.
3. Start Local Server.
4. Make sure the endpoint is available at http://127.0.0.1:1234/v1.

4. Connect an MCP client

Connect LINZA to Claude Desktop, Cursor, OpenCode, or any MCP client:

{
  "mcpServers": {
    "linza": {
      "command": "linza-mcp",
      "env": {
        "LINZA_VAULT": "/absolute/path/to/workspace-or-vault",
        "LINZA_EMBED_PROVIDER": "lmstudio",
        "LINZA_EMBED_URL": "http://127.0.0.1:1234/v1",
        "LINZA_EMBED_MODEL": "your-embedding-model-name",
        "LINZA_TOOL_SURFACE": "default"
      }
    }
  }
}

VS Code / Copilot MCP uses servers:

{
  "servers": {
    "linza": {
      "type": "stdio",
      "command": "linza-mcp",
      "env": {
        "LINZA_VAULT": "/absolute/path/to/workspace-or-vault",
        "LINZA_EMBED_PROVIDER": "lmstudio",
        "LINZA_EMBED_URL": "http://127.0.0.1:1234/v1",
        "LINZA_EMBED_MODEL": "your-embedding-model-name"
      }
    }
  }
}

5. Check the setup

linza-mcp --version

Then ask the agent:

Check LINZA with agent_workspace(action="doctor").
Index the folder and show the first 3-5 review cards.

Optional Docker Run

Docker is not required, but the repository includes a small image for isolated stdio runs:

docker build -t linza-mcp .
docker run --rm -i `
  -v /absolute/path/to/workspace-or-vault:/data/vault `
  -e LINZA_EMBED_PROVIDER=lmstudio `
  -e LINZA_EMBED_URL=http://host.docker.internal:1234/v1 `
  -e LINZA_EMBED_MODEL=your-embedding-model-name `
  linza-mcp

Use host.docker.internal only when the embedding server runs on the host machine. Otherwise pass the embedding API URL that is reachable from inside the container.

---

Embeddings

Embeddings are not decoration; they are the quality of LINZA's sight. The main path is simple: a local model in LM Studio and an MCP server next to your folder.

• lmstudio is the recommended local setup. Use it when you want good semantic search, topic maps, and links without cloud calls.
• ollama is a local Ollama setup.
• openai is any OpenAI-compatible endpoint with /embeddings.

Example LM Studio environment:

$env:LINZA_EMBED_PROVIDER="lmstudio"
$env:LINZA_EMBED_URL="http://127.0.0.1:1234/v1"
$env:LINZA_EMBED_MODEL="your-embedding-model-name"

If you switch embedding provider or model dimension, run a full reindex. Vectors from different models live in different spaces and should not be mixed.

---

First Output Example

Agents usually start with agent_workspace(action="doctor") or guide_next_steps(language="en"). A human should see a short status, not a raw JSON wall:

LINZA is ready

Material:
- 42 notes indexed
- 3 incoming artifacts waiting for review
- sidecar: .linza/linza.db

Next step:
1. Review discovered domains
2. Accept, rename, or skip 3-5 cards
3. Nothing is written without dry-run/apply

Example card:
Proposal: connect "Retrieval Quality Note" and "Source Policy"
Why: shared vocabulary, review-flow references, nearby chunks
Write impact: none yet; accepting records a sidecar relation

---

MCP Tools

LINZA exposes a compact default surface of 15 tools. Normal operation starts from agent_workspace or guide_next_steps, not from a raw tool list.

Tool	Purpose
agent_workspace	One facade for map, ingest, review, teach, grow, connect, memory search, context export, calibr, and doctor
guide_next_steps	Show the next safe step
index_all	Index the Markdown folder into .linza/linza.db
search	Semantic and lexical search
read_file	Read an exact file from the vault
get_stats	Quick sidecar counters
scan_vault	Read-only folder diagnostic
build_review_apply_queue	Build review cards with stable rq-* IDs
approve_review_queue_items	Dry-run or apply selected cards
list_approved_items	Show accepted sidecar items
explain_node	Explain one node: links, bridges, context
explain_relationship	Explain a possible relation between two nodes
who_depends	Show dependencies and neighbors
show_flow	Find a route or flow between nodes
create_context_pack	Build a compact context pack for an agent

agent_workspace(action="teach") selects seed cards. grow returns a preview with selected_rules: why each card entered the batch. The autonomy model is simple: teach with examples, grow in preview, apply in small reviewed batches.

The advanced surface exists for development and debugging:

$env:LINZA_TOOL_SURFACE="advanced"

See the full Tool Catalog.

---

Artifact Inputs

Supported inputs:

• pasted text;
• local .md, .txt, .json;
• local .docx, .xlsx;
• local .pdf when pypdf or PyPDF2 is installed.

Logs do not need a special file format. Paste them as text or save them as .txt.

LINZA does not open web pages by itself. The agent uses its own browser/web-fetch tool, extracts readable text, and passes it to LINZA as an artifact, for example source_kind="web_article" or source_kind="browser_capture".

Imported text is analysis material, not an agent instruction. This is the basic prompt-injection boundary: instructions inside an article, log, chat, or PDF are not executed. Memory, rules, and YAML appear only after review.

---

Safety Model

LINZA is a local review-gated sidecar:

• indexing, analysis, and import do not change source note bodies;
• raw artifacts stay in local SQLite storage;
• generated reports write only under .linza/reports;
• context packs write only under .linza/context-packs;
• visible YAML edits are compact and require review/apply;
• hierarchy, causal links, memory, calibr lessons, and approvals stay in the sidecar until the human asks for export.

LINZA is not a browser automation server, cloud memory, or an autopilot that silently rewrites rules, skills, memory, or notes.

Stability

0.1.3 is an alpha MVP. The safety contract is meant to be stable: source note bodies are not rewritten by indexing, artifact ingest, search, map, or grow preview. Low-level advanced tools and internal module boundaries may still change while the server is being polished.

---

Agent Pack

The repository includes a portable skill for agents:

agent-pack/skills/linza-operator/SKILL.md
agent-pack/skills/linza-operator/references/workflows.md
agent-pack/skills/linza-operator/references/safety-policy.md
agent-pack/skills/linza-operator/references/tool-audience.md

The skill tells an agent what to show humans, when to use agent_workspace, how to handle URLs through an external browser/web-fetch tool, and why apply actions must be dry-run or exact-ID gated.

---

Example And Verification

The synthetic private-safe example pack lives in:

examples/sample-vault/
examples/artifacts/
examples/expected/

Run the full regression suite:

python -m unittest

Run one specific test:

python -m unittest test_agent_workspace.AgentWorkspaceTests.test_examples_sample_pack_runs_end_to_end

---

Environment Variables

Variable	Description
LINZA_VAULT	Path to the Markdown folder
LINZA_EMBED_PROVIDER	lmstudio for the recommended local setup; openai or ollama are also supported
LINZA_EMBED_URL	Embeddings API URL
LINZA_EMBED_MODEL	Embedding model name
LINZA_EMBED_KEY	Optional key for an OpenAI-compatible embeddings API
LINZA_BRIDGE_THRESHOLD	Semantic bridge threshold; default 0.55
LINZA_MAX_BRIDGE_PAIRS	Maximum note pairs for semantic bridge rebuilds; default 1000000, 0 disables the guard
LINZA_DEFAULT_PROFILE	Default search profile name; default general
LINZA_TOOL_SURFACE	default (15 tools) or advanced
LINZA_LANGUAGE	Human-facing guide language: auto, en, or ru

---

_{Cosines do not converge because everything is the same. They converge when meaning finds its angle.}