What is Sombra?

Sombra is a research library for developers and AI agents. It lets you save web pages, organise them into collections, distil what matters from them, and feed all of it to your AI coding tools via MCP — in a single, coherent workflow.

The tagline: Your AI's research library.

The problem it solves

Knowledge workers — especially developers — accumulate enormous amounts of reference material as they work: documentation pages, blog posts, migration guides, architecture writeups, competitor pages, research papers. Today this material is scattered across browser tabs, bookmarks folders, Notion pages, and local markdown files. It exists in formats that are useless to AI agents.

When you sit down with Claude Code or Cursor to tackle a complex task, your agent doesn't know what you know. It can't read what you've saved. You end up re-pasting the same docs, re-explaining the same context, over and over. This is the context gap.

Sombra closes it.

The three-step workflow

1. Save

Install the Chrome extension and save any web page with one click. Sombra archives the clean, readable content — stripping ads, nav, and boilerplate — and stores it permanently. If a page goes offline or changes, your saved version remains intact.

2. Organise

Group saved content into Collections — thematic containers you define. A collection might be "Pedestal migration docs," "Competitor pricing pages," "Belgian tax regulation," or "React 19 upgrade notes." Collections are the unit of context: everything you've gathered around a topic, in one place.

3. Distil and build

Write Collection Context — a distilled, structured summary of what matters in a collection. Not a dump of raw content, but the extracted signal: key APIs, breaking changes, implementation patterns, architectural decisions. This is the layer that makes collections genuinely useful to AI agents. Your AI doesn't need to read 40 pages of docs; it needs the 500 tokens that actually affect what it writes.

Connect Sombra to any MCP-compatible AI tool with a single command:

claude mcp add --transport http --scope user sombra https://sombra.so/mcp

From that point, Claude Code, Cursor, or any MCP client can browse your collections, read saved pages, and use distilled context — directly, without copy-pasting anything.

AI-assisted collection building

You don't have to build collections manually. Because Sombra is accessible via MCP, your AI assistant can do the research legwork for you — finding relevant pages, saving them, organising them into a collection, and writing a distilled context — all in a single conversation. Ask it to research a topic, and come back to a structured, ready-to-use knowledge base.

Core features

URL saving & web clipping One-click saving via Chrome extension. Clean markdown extraction. Permanent archive — no dead links. Original URL preserved with full metadata.

Collections Organise saved content into named, themed groups. Collections are the primary unit of context for AI agents. Move artifacts between collections, rename and describe them, keep your knowledge structured as projects evolve.

Collection context / distillation Write a structured summary for any collection — or have your AI write it. This is the differentiating layer. Good distillation is dense, technical, and actionable — preserving code examples verbatim, surfacing the key facts an agent needs without forcing it to read everything.

MCP integration Sombra exposes a remote MCP server. Connect once, and every MCP-compatible client — Claude.ai, Claude Desktop, Claude Code, Cursor — can access your full library. No local setup, no file management, no repeated copy-paste.

Note artifacts Create freeform markdown notes directly in Sombra. Useful for in-progress thinking, distillation drafts, project summaries, or anything that doesn't have a URL but belongs with your other saved material.

Public sharing Share any collection via a public URL. Shared collections show your distilled context, your notes, and your sources as cited references — title, link, and your annotation. Extracted page content stays private, both out of respect for the original authors and to keep shared collections focused on your thinking rather than other people's words.

Use cases

Developers: API and library research

You're integrating HubSpot's UI component library, or wiring up the Java Elasticsearch client, or navigating AWS CloudFormation. Instead of juggling twenty browser tabs, you save the relevant docs into a collection as you read. Your AI distils the key interfaces and gotchas into a context note. From that point on, every coding session starts with your agent already knowing the API — no re-pasting, no re-explaining.

Cybersecurity: vulnerability intelligence

A new CVE drops with multiple advisories, vendor patches, and researcher writeups scattered across different sites. Save them all into a single collection, distil the key details — affected versions, attack vectors, mitigations — and share it as one coherent unit with your team or your AI. Everything in one place, permanently archived even if the original sources change.

Sales: prospect research

Before a call, task your AI with building a collection on the prospect — their website, recent press coverage, LinkedIn posts, funding announcements, job listings. It saves, organises, and distils the relevant details into a briefing. You walk into the meeting already knowing their priorities, their stack, and what's changed in the last quarter.

Journalism: source and story collation

You're working a long-form story across weeks of research. Save every source — articles, public records, official statements, background reading — into a collection as you find them. Your AI can help you identify gaps, surface contradictions across sources, and draft from a curated base rather than hallucinating from training data.

Competitive intelligence: living briefs

Build a collection tracking a competitor — their pricing page, product changelog, job postings, press releases. Refresh it as things change. The distilled context becomes a living brief your AI always has access to: current positioning, recent moves, strategic signals. No more stale summaries.

Legal and compliance: matter research

Collate the relevant regulation, case references, and guidance documents for a matter into a single collection. Your AI drafts against a curated, citable source base rather than general training knowledge — reducing hallucination risk and keeping everything traceable back to primary sources.

Academic research: literature and notes

Save papers, preprints, and reference articles into a collection as you build a literature review. Add your own notes as artifacts. Distil the state of a field — key findings, open questions, methodological debates — so your AI can help you write with genuine grounding rather than confabulated citations.

Writers and creators: persistent AI briefing

Save your style guide, audience research, tone notes, and reference examples into a collection. This becomes your AI's permanent briefing — it knows your voice, your readers, and your conventions every time you start a new piece. No more re-explaining who you're writing for at the top of every conversation.

What makes it different

Most bookmarking and knowledge tools are built for human recall — for you to find things later. Sombra is built for AI recall — for your agent to have the right context at the right moment.

The distillation layer is the key differentiator. Nobody else does "save web pages → organise into projects → write distilled technical context → feed to coding agent via MCP" as a unified, hosted flow through a single interface. Local markdown files (like Will Larson's "datapacks" concept) get partway there, but require manual file management, lack web extraction, and can't serve multiple AI clients simultaneously.

When you share a collection publicly, what you're sharing is your research and your thinking — the sources cited, the context distilled, the notes written. That's the valuable part. Sombra is hosted, persistent, and MCP-native. Your library is available everywhere your AI is.