DocsAgent MCP server

DocsAgent MCP — Zotero, Obsidian & Apple Notes for AI agents 📚⚡

DocsAgent gives AI agents instant, private access to your personal knowledge base. docsagent is the spec-driven MCP (Model Context Protocol) server that lets any AI agent — Claude Desktop, Cursor, Cline, Qwen Code, or any MCP client — search, read, and write your Zotero library, Obsidian vault, and Apple Notes (macOS) through a resident C++ search engine. BM25 full-text search + query-ranked passage retrieval over 1,000+ PDFs at ~15 ms, fully local (RAG-ready knowledge base).

• 🔒 Local-first & private — the engine reads your Zotero library, Obsidian vault, and Apple Notes store directly on your machine. Your notes and PDFs never leave it.
• 🗂 Three sources, one query — Zotero items/annotations/notes, Obsidian vaults, and Apple Notes (macOS only); mixed search fuses all three by reciprocal rank.
• ⚡ Native C++ search core — inverted-index BM25 + passage ranking, millisecond lookup, low memory footprint (160–227 MB for a 1,500-paper library).
• 🧩 8 MCP tools — 5 read + 3 write, with JSON-schema validated arguments, token budgets, result dedup, and a three-layer write safety gate.
• 🌐 Two transports — stdio for local MCP clients, Streamable HTTP for remote deployment (origin checks, API-key / OAuth 2.0 token introspection, per-request RBAC, /health probe).
• 🐍 Two shells, one core — this TypeScript package and a feature-equal Python wrapper ship the same tools over the same JSON-RPC contract.

---

Architecture

MCP Client (Claude Desktop / Cursor / Cline / Qwen Code / any MCP host)
        │  stdio (local)   or   Streamable HTTP  /mcp  (remote)
        ▼
MCP shell  ← this package (@docsagent/mcp-zotero / docsagent-mcp-zotero)
   · tool schemas (spec-driven), argument validation, token budget, dedup
   · write orchestration via the Zotero local API, write safety gate, RBAC
   · group-library sync via the Zotero Web API
        │  JSON-RPC 2.0 over HTTP ({coreHost}:{httpPort}/rpc, cpp-httplib)
        ▼
DocsAgent Core (resident C++ engine)
   · reads ~/Zotero/zotero.sqlite + storage/, an Obsidian vault, and (macOS) the
     Apple Notes store — directly on your machine
   · builds & serves one index per source (BM25 + passage ranking); mixed search
     (source "all") fuses them by reciprocal rank (RRF, k=60)

The shell never spawns the core during tool calls and never touches your source files. The core runs as a background service and stays available across MCP client restarts. Tool schemas and error codes: spec/.

---

Quick Start

1. Start the core

npx @docsagent/docsagent start           # spawn the bundled core for your platform
npx @docsagent/docsagent status          # pid / endpoint / version

Python shell (same verbs, under the core subcommand):

pip install ./python                      # build the wheel locally (PyPI upload pending)
docsagent-mcp core start

(core stop / core restart also available. The core indexes every configured source — your Zotero data directory, your Obsidian vault, and (macOS) Apple Notes — and serves JSON-RPC on http://0.0.0.0:23120/rpc.)

2. Configure your MCP client

Claude Desktop / Cursor / Cline / Qwen Code (mcpServers):

{
  "mcpServers": {
    "docsagent": {
      "command": "npx",
      "args": ["-y", "@docsagent/docsagent"]
    }
  }
}

Python shell (same tools, same contract, installed from this repo) — mcpServers:

{
  "mcpServers": {
    "docsagent-zotero-py": {
      "command": "docsagent-mcp"
    }
  }
}

Install the Python package first (step 1) so docsagent-mcp is on your PATH.

On startup the shell connects to the core, loads sources, and checks index status. If the core is not running it fails fast with startup instructions — it never spawns anything.

---

MCP Tools (external API)

8 tools, 5 read + 3 write. Schemas are the single-sourced contract in spec/tools/*.json (mirrored into both packages); arguments are validated before handlers run and failures map to typed docsagent error codes.

list_sources

Every searchable source with capabilities, supported targets/includes/browse modes, filters, and document counts. Call this first.

Sources today: zotero (targets items, annotations, notes; collections/tags browse), obsidian (a vault — notes as items, folders/tags browse), and apple-notes (macOS only — the engine omits the source on other platforms). The vault is found at ~/Documents/Obsidian Vault (override with DOCSAGENT_OBSIDIAN_VAULT); Apple Notes is read from the system Notes store (override with DOCSAGENT_APPLE_NOTES_DB).

search

Cross-entry search over one source (items, annotations, and notes).

Parameter	Type	Notes
query	string	plain keywords or phrases; required unless mode=grep
mode	relevance | grep	BM25 relevance by default; grep scans for a literal pattern without tokenization or ranking
pattern	string, required when mode=grep	literal string to scan for
caseSensitive, wholeWord, maxMatches		mode=grep only: ASCII case folding, [A-Za-z0-9_] word boundaries, hit cap total
target	"items" | "annotations" | "notes" or array	default items
depth	ids | snippets | full	snippets by default (BM25-ranked passages)
filters	object	tags, yearFrom/yearTo, itemType, authors, colors, containerId, titleContains
k, snippetsPerResult, max_tokens	numbers	ranking depth and token budget (mode=grep: max documents, max hit windows per document)

Returns results[] with global ids (zotero:KEY, obsidian:<path>, apple-notes:<uuid>), titles, relevance, snippets; multi-target searches group by target. In mode=grep each result carries matchCount and snippets[] hit windows (hits[] with line/column/offset, and meta hits tagged with field), relevance is 0, and the response adds totalMatches. Results are deduped (id, then normalized title + year) and packed under a token budget.

search runs against config.defaultSource (default zotero). Mixed search is a core capability: pass source: "all" to the core's search / grep methods and every source is ranked independently, then fused with reciprocal rank fusion (k=60) — each hit is tagged with its source. BM25 scores are not comparable across corpora, so fusion is rank-based.

get_content

Read one entry. mode=passages (query-ranked passages, k) or mode=fulltext (offset pagination with nextOffset). Notes return their body with tags and metadata.

get_metadata

include: metadata, abstract, annotations, notes, citation (bibtex / csljson / formatted via citationFormat/citationStyle). Notes are packed under the token budget.

list_library

Browse modes: collections (drill-down via parentId), items (by containerId), tags, saved_searches, standalone_notes. Browse modes are per source — Zotero exposes all of these, while Obsidian and Apple Notes expose folders (drill-down) / tags / items.

Write tools (three-layer safety gate)

Tool	What it does	Key arguments
import_item	Import local PDFs or resolve DOI / ISBN / arXiv IDs (via the Zotero translation server); optional autoClassify suggests collections	paths | identifiers, containerId, autoClassify, confirmed
add_note	Add a Markdown child note to an item (converted to Zotero note HTML), with orphan verification and rollback	id, content, tags, confirmed
batch_modify	Bulk add_to_collection / remove_from_collection / add_tags / remove_tags on up to 200 items in batches of 50	action, ids, containerId, tags, confirmed

Write tools target Zotero sources only — the Obsidian and Apple Notes sources are read-only.

Write safety gate (spec/algorithms/write-gate.md): layer 1 write tools are not registered unless enableWrites=true; layer 2 confirmed=false returns a preview and consumes no rate-limit quota; layer 3 confirmed writes consume a per-hour rate limit (default 30/h). Anything above 20 items in batch_modify additionally reports requiresConfirmation in the preview.

---

Engine performance

The C++ engine powers PapersGPT — the same index and retrieval stack ships in this MCP server. Benchmark on a real Zotero installation (full write-up):

Metric	Mac (Intel i9)	Windows VM (4C8G)
Library size	1,506 PDFs (4.5 GB on disk)	500+ PDFs
Index build time	141 s	a few seconds
Memory (agent process)	227 MB	160 MB
Average retrieval latency	~15 ms	~15 ms

• Indexing cost scales roughly linearly with library size; retrieval latency stays constant — a 10,000-paper library (~30 GB) indexes in about 15–20 minutes, and everyday search stays at ~15 ms.
• For comparison: a typical web page load takes 1,000–3,000 ms; a blink of an eye is 100–150 ms. PapersGPT answers in ~15 ms, fully offline.
• Privacy: your library never leaves your machine.

Configuration

Config lives at ~/.docsagent/config.json (or $DOCSAGENT_CONFIG) — one file shared by the JS shell, the Python wrapper, and the C++ core. Validated against spec/config.json.

Key	Default	Description
coreHost	0.0.0.0	Address the core binds and the shell dials
httpPort	23120	Core HTTP port (POST /rpc)
coreBinary	""	Optional explicit path to the core binary
zoteroDataDir	~/Zotero	Zotero data directory
zoteroApiUrl	http://localhost:23119/api	Zotero local API (write orchestration)
zoteroGroups	[]	Group libraries to sync from zotero.org
enableWrites	false	Register the three write tools
writeRateLimitPerHour	30	Confirmed-write rate limit
maxTokensPerTool	4000	Token budget per tool result
defaultSource	zotero	Source the shell searches and resolves id prefixes in (zotero, obsidian, apple-notes)
transport	stdio	stdio or streamable-http
httpListenAddr	0.0.0.0:8080	Listen address for streamable-http (/mcp)
authMode / authConfig	none	api-key or oauth2 (RFC 7662) + allowedOrigins
rbacRoles	{}	role → allowed tool names (per-request RBAC on HTTP)
logLevel	info	debug / info / warn / error

---

Distribution

Channel	Package	Bundled core	Size
npm (JS/TS shell)	@docsagent/docsagent	all platforms in bin/	~70 MB tarball
PyPI (Python shell)	docsagent-mcp (pip install ./python)	same binaries in the wheel	~65 MB wheel

Bundled core platforms: macOS universal (Intel + Apple Silicon), Windows x64 (x86_64), Linux x64 (x86_64) — Linux ARM is not supported. Source availability: Zotero and Obsidian work on every platform; Apple Notes is macOS-only (list_sources omits it elsewhere).

Both shells read the same config and talk to the same core — pick either (or both) as your MCP distribution channel. Core lifecycle (start / stop / restart / status) is available from both CLIs.

Links

• spec/ — the single-sourced contract: 8 tool schemas, 23 JSON-RPC methods, error codes + suggested calls, config schema
• PapersGPT search performance benchmark
• Zotero · Model Context Protocol

License

Apache-2.0