# Levirge > Levirge provides shared knowledge and web research services for AI agents. ## Connect - [Levirge Brain MCP definition](https://levirge.com/brain/mcp.json): Streamable HTTP server definition (JSON). - [Levirge Search MCP definition](https://levirge.com/search/mcp.json): Streamable HTTP server definition (JSON). - [Both products](https://levirge.com/mcp.json): one definition file with both servers. ## Products - [Levirge overview](https://levirge.com/index.md): Platform overview and product architecture. - [Levirge Brain](https://levirge.com/brain.md): MCP-native shared memory, reusable skills and agent handoffs. - [Levirge Search](https://levirge.com/search.md): Search, retrieval and targeted extraction from public web sources. ## Levirge Brain - [Brain FAQ](https://levirge.com/brain/faq.md): Differences from RAG and vector stores, clients, auth, privacy, hosting and behaviour. - [Register interest in Brain](https://levirge.com/brain/register.md): Plan-level interest registration for Levirge Brain. - [Brain docs](https://levirge.com/brain/docs.md): Index of Brain documentation — install, tools, roadmap. - [Brain MCP tool reference](https://levirge.com/brain/docs/mcp-tools.md): Human-focused reference for Brain's 19 MCP tools and the terms they use. - [Brain roadmap](https://levirge.com/brain/docs/roadmap.md): Shipped with ship dates; in build and designed-not-started without. - [Brain on-premise](https://levirge.com/brain/docs/on-premise.md): Self-hosted deployment — status, what it involves, and the waitlist. - [On-Premise requirements](https://levirge.com/brain/docs/on-premise/requirements.md): Hardware, services and the questions that size a self-hosted Brain deployment. - [On-Premise model options](https://levirge.com/brain/docs/on-premise/model-options.md): Brain's four inference models — embeddings, summarising, enrichment, reranking — and where each can run. - [Examples](https://levirge.com/brain/examples.md): Worked Brain examples — measured before/after runs. - [Dense results](https://levirge.com/brain/examples/dense-results.md): Measured before/after — eight fragments to synthesise, or one cited summary consolidated from eleven sources. - [Closed-loop handoff](https://levirge.com/brain/examples/closed-loop-handoff.md): A real multi-agent round trip — benchmark findings to merged fixes to verification. - [Install Brain in Claude Code](https://levirge.com/brain/docs/install/claude-code.md): Marketplace install, OAuth sign-in, and the /brain:* commands. - [Install Brain in Claude Desktop](https://levirge.com/brain/docs/install/claude-desktop.md): Marketplace installation without a terminal. - [Install Brain in Codex](https://levirge.com/brain/docs/install/codex.md): Marketplace command, or MCP server via config.toml. - [Install Brain in Obsidian](https://levirge.com/brain/docs/install/obsidian.md): Two-way sync between an Obsidian vault and the shared knowledge base. - [Connect any MCP client to Brain](https://levirge.com/brain/docs/install/mcp.md): Endpoint, OAuth or bearer token, and a read-only verification call. ## Levirge Search - [Register interest in Search](https://levirge.com/search/register.md): Plan-level interest registration for Levirge Search. - [Pricing](https://levirge.com/search/pricing.md): Plan ladder — unlimited search and open fetch on every plan; users and full-browser (stealth) capacity per tier. - [Receipts](https://levirge.com/search/receipts.md): Dated, verbatim native-fetch vs Levirge fetch comparisons — the running evidence table. - [Examples](https://levirge.com/search/examples.md): Seven worked retrieval and research scenarios. - [FAQ](https://levirge.com/search/faq.md): Supported sources, clients, routing, caching and failure handling. - [Install in Claude Code](https://levirge.com/search/install/claude-code.md): Marketplace install — two commands. - [Install in Claude Desktop](https://levirge.com/search/install/claude-desktop.md): Marketplace installation without a terminal. - [Install in Codex](https://levirge.com/search/install/codex.md): Marketplace command, or MCP server via config.toml. - [Install in OpenCode](https://levirge.com/search/install/opencode.md): Remote MCP server — one endpoint. - [Install in pi.dev](https://levirge.com/search/install/pidev.md): Remote MCP server — one endpoint. ### Examples - [Schema-first discovery](https://levirge.com/search/examples/schema-discovery.md): Inspect a JSON shape or Markdown outline before selecting content. - [Oversized source](https://levirge.com/search/examples/oversized-source.md): Reduce a 1.26 MB GitHub response to 28 KB with server-side selection. - [Quiet failure](https://levirge.com/search/examples/quiet-failure.md): Report challenge pages and failed retrieval routes as explicit errors. - [Queued reads](https://levirge.com/search/examples/queued-reads.md): Queue multiple URLs and collect completed, pending and failed results. - [Blocked portal](https://levirge.com/search/examples/blocked-portal.md): Compare an ordinary 403 response with browser-backed retrieval. - [Search and direct fetch](https://levirge.com/search/examples/fast-path.md): Receive ranked search results and size-capped Markdown. - [Release digest](https://levirge.com/search/examples/release-digest.md): Build a scheduled release summary from selected GitHub API fields. ## Company and policies - [Contact](https://levirge.com/contact.md): Discuss a use case or workspace. - [Trust](https://levirge.com/trust.md): Index of service governance documents. - [Security](https://levirge.com/security.md): Workspace isolation, token handling and operational safeguards. - [Privacy](https://levirge.com/privacy.md): Data collection, retention, model lanes, subprocessors and your rights. - [Terms](https://levirge.com/terms.md): Early-access terms: accounts and agents, acceptable use, data ownership, fees, liability, disputes. - [Blog](https://levirge.com/blog.md): Essays on agent memory, knowledge capture, retrieval and distillation. - [Capture: stopping the leak at the point of creation](https://levirge.com/blog/capture-stopping-the-leak.md): Part 2 of 4 - capture has to happen at the moment of creation, or not at all. - [Every chat session ends with a small layoff](https://levirge.com/blog/every-session-ends-with-a-small-layoff.md): Part 1 of 4 - the knowledge a working session creates, and what survives it.