Skip to content

Librarium

Multi-provider deep research CLI — fans out queries to multiple search/AI APIs in parallel

Install

$claude mcp add librarium -- librarium mcp

Install command taken from the project's README or registry entry. Always review code before running it.

Librarium

Evidence-aware, multi-provider research for people and agents.

= 22.12" />

Website · Quick start · Catalog · CLI · Agents

What Librarium does

Librarium plans a research request against explicit provider profiles, executes the selected work, and records what happened. It can combine web search, grounded answers, consumer-surface observations, and durable research jobs. It does not turn agreement into proof. Results preserve their collection and access provenance so a caller can judge what each result means.

Version 2 separates the product into three import boundaries:

Import Use it for It never does on import
librarium Worker-safe schemas, the public catalog, and pure config/contract utilities Read the host, load adapters, access credentials, start the CLI, or write files
librarium/core Advanced injected planning, transport, coordination, and execution ports Supply caller-owned runtime dependencies, construct concrete provider adapters, or create a global registry
librarium/node Explicit Node config, credential, trusted custom-provider, and canonical validation services Provide a turnkey built-in-provider runner or a portable persistence layer

These library boundaries are deliberately compositional. A caller using librarium/core owns the concrete adapters, credential context, clocks, coordination store, artifact services, and other runtime dependencies required by the ports it invokes. librarium/node adds host-specific building blocks; it does not assemble the built-in catalog into a high-level runner.

The librarium CLI and its MCP stdio server are the complete application interfaces for research runs. Node package installs require Node.js 22.12 or newer. Standalone and Homebrew binaries include their own runtime.

Quick start

npm install -g librarium

# Configure credentials and select providers interactively.
librarium init

# Inspect the exact offline preflight plan before anything runs.
librarium plan "PostgreSQL connection pooling"

# Run the default quick workflow and wait for its providers concurrently.
librarium run "PostgreSQL connection pooling"

# Or make an evidence-bounded answer from the quick workflow.
librarium answer "What changed in PostgreSQL 17?" --verify

Use librarium doctor before a paid run to check configuration and credential presence offline without loading configured custom provider code. Connectivity is not tested unless you explicitly run librarium doctor --live, which loads trusted custom providers, makes provider network requests, and may incur charges. librarium run --json sends only JSON to standard output; progress and diagnostics go to standard error. New requests default to the quick workflow in sync mode. Pass --group, --providers, or --mode to override those defaults; saved configuration preferences are also honored.

V2 catalog

The v2 catalog has 34 built-in providers and 41 implemented public profiles. A profile is the unit of selection and provenance: provider_id/profile_id. A provider can expose more than one profile, such as exa/search and the durable exa/research profile. Adapter IDs are a Node implementation detail; do not persist an internal adapter ID as public v2 configuration.

Workflows

Only these four names are built in:

Similar plugs

Turn any codebase, with its docs, SQL schemas, configs, and PDFs, into a queryable knowledge graph. A /graphify skill for Claude Code, Cursor, Codex, and Gemini CLI: local…

Why use many token when few token do trick. Viral skill + proxy for coding agents that cuts 65% of tokens by talking like a caveman.

The 1st Collaborative AI Design Agent Workspace. Local-first and open source. Create prototypes, websites, apps, dashboards, presentations, images, and videos with your team…