Contexelcontext engine

Context engine for AI coding agents

Stop pasting the same context into your AI agent.

Every session you paste the same background: how the code works, which rules win, what not to touch. Contexel keeps it in one place and hands your agent only the parts that fit the task, with a source on every one. Your own Markdown files stay the source of truth. Sign in with your browser and connect your agent in one line.

Open the console See how it works
claude mcp add --transport http contexel https://contexel.ai/mcp

Sign in to the console to get your exact connect command with your token. Works with Claude Code, Cursor, and any MCP client. MCP is the open standard agents use to plug into tools.

get_context hostedbeta
› get_context({ task: "add rate limiting to /login" })
3 cards · 1,180 tokens · project: work
auth/rate-limit.md#policy top rule · team
api/login.md#handler source · git@a1c…
conventions.md#error-handling source · org
Hosted for your team · A source on every answer · Fits your agent's size limit · Same files in, same context out
Works with Claude Code, Cursor, and any MCP client.

The problem

More room in the prompt did not fix the pasting.

Agents drown in stale CLAUDE.md files, notes that leak from one project into another, and rules that contradict each other with no way to tell which one wins. So you paste the right background by hand, every session. Contexel turns that into a question your agent asks, instead of a ritual you repeat.

How it works

Ask once. Get the right context back.

1

Sign in and connect your agent.

Open the console, then point Claude Code, Cursor, or any MCP client at your workspace address. You do this once.

2

Ask in plain words.

Say something like "what do we know about rate limiting?" Your agent knows what to do: the tools have plain-English names, and the service briefs your agent the moment it connects. Nothing to paste. If it is ever unsure, a built-in help tool explains what to ask.

3

It gets back only what matters.

Sorted to the right project, with conflicts already resolved, a source on every piece, and trimmed to fit your agent's size limit.

When you would reach for it

Situations you'll recognize.

Three sessions in, it asks which package manager you use. Again.

You are deep in a refactor and your agent has forgotten, again, where the migrations live, why the auth middleware is shaped the way it is, and which package manager you use. So you paste the same three paragraphs into the prompt one more time. With Contexel those facts live as small Markdown notes. Your agent calls get_context with the task and gets back only the ones that fit, inside its size limit, with the source stamped on each. When it learns something new mid-session it saves a note with observe, and you approve anything lasting. The third session starts as sharp as the first one ended.

Every engineer runs an agent. Every agent needs the same house rules.

You are on a platform team where the error-handling conventions, the deploy checklist, and the never touch prod config rule live in a dozen slowly diverging CLAUDE.md copies. With a shared Contexel workspace there is one copy. Teammates sign in with Google, GitHub, Microsoft, or SSO, and each agent points at the same address with a token you mint or a browser sign-in. When rules disagree, precedence decides the same way for everyone (company rule first, then team, then your own), and the answer says which rule won and why. Change the rule once and every agent has it on the next call.

New repo, and the agent spends its first hour rediscovering what the last person already knew.

You have been pointed at a codebase you did not write. Bring the docs you already have: its existing CLAUDE.md maps into routed context section by section, and the repo's Markdown docs can be mirrored in read-only with the GitHub connector. The moment your agent connects, the service briefs it on what Contexel is and how to ask. A built-in help tool covers the rest, and status shows what is loaded. Its first get_context call returns the maintainers' knowledge, with a source pointing back to where it came from.

Five clients. Client A's architecture must never surface in client B's session.

You are consulting across five clients, and one big notes pile with search is a confidentiality incident waiting to happen. In Contexel each client is its own project (a realm), kept in a separate store rather than behind a filter, so a question is answered from exactly the project you name. Tokens are scoped per project and fail closed: access to one never grants access to another. And when you genuinely want to compare notes across two clients, you ask both by name, and every note comes back labeled with where it lives. Kept apart by design, not by discipline.

The security review asks where the agent's context lives and who can read it.

You are rolling out agents in an org where "it is in a vector database somewhere" will not pass review. A Contexel workspace answers with specifics: encryption you can turn on so your documents and search index are encrypted at rest under a key dedicated to your workspace, tokens scoped per project that fail closed, every read, write, and denial in a tamper-evident audit log, and writes that land only through human-gated review as git commits with a source. On the key itself you get a real choice. By default Contexel holds it so it can run indexing and search for you. Or switch to a key only you hold and supply it each session; it lives in your workspace's memory only while unlocked, and when the workspace is locked, Contexel has no way to decrypt your documents or your search index. You control all of it from the console.

What it does

Context you can trust.

Separate projects stay separate

Work, personal, one space per client. Each is a separate project (called a realm in the app). Context from one never leaks into another. They are kept apart by design, not by a filter that could slip.

It knows which rule wins

When two instructions disagree, Contexel picks the one that should win (company rule first, then team, then your own) and shows you why.

A source on every answer

Every piece of context says where it came from. Nothing is anonymous, and nothing is dropped without telling you.

You are never locked in

Your content is plain Markdown that you own. The search index is built from it and can be thrown away and rebuilt, so there is no proprietary format to escape. Every workspace is walled off, and you can turn on encryption at rest.

Same files in, same context out

Run it twice over the same files and you get the same result, byte for byte. No drift, no surprises.

It fits the space your agent has

You set a size limit (a token budget: how much text an AI can read at once). Contexel packs the most relevant context to fit, and tells you if it had to leave anything out.

Nothing new to learn

Every tool has a plain-English name, your agent is briefed automatically when it connects, and a built-in help tool answers "what is this and what do I say?" If you can describe your task, you can use it.

It remembers. You stay in control.

Your agent can save a quick note (observe) or draft a lasting change (propose) that a person approves (review). Auto-save is off by default, every saved change records its source, and secrets are caught before anything is written.

Start with the notes you already have

Point Contexel at a flat instruction file like CLAUDE.md and it maps cleanly into context sorted by project. No rewrite, no new format. Markdown stays the source of truth.

Bring in outside sources

A GitHub repo of Markdown can be mirrored in read-only, and a meeting transcript can be brought in as review-queued drafts. Everything passes secret redaction first, and distilled content never lands without approval.

Two ways for agents to connect

Mint a token in the console, or let your agent sign in with your browser identity (OAuth 2.1) and grant it access yourself, no operator needed. Tokens are scoped per project and fail closed.

Encryption you control

Turn on encryption and your documents and search index are encrypted at rest under a key dedicated to your workspace. Contexel holds the key by default, or you switch to a key only you hold. Rotate or export the managed key from the console.

Built for teams

One shared context service for the whole team.

For your teammates

Sign in and go.

Sign in with your browser using Google, GitHub, Microsoft, or your company SSO. Every workspace is walled off from the others, and you can turn on encryption at rest. No install, no setup.

Open the console
Custodial today. See "Straight talk" below.
For your agents

Connect in one line.

Point Claude Code, Cursor, or any MCP client at your workspace address. Agents connect with a token you mint, or with a browser sign-in (OAuth 2.1) where it is turned on, then call get_context, observe, and propose.

See the connect steps

Encryption at rest

You choose who holds the key.

Turn on encryption and your documents and your search index are encrypted at rest under a key dedicated to your workspace. You decide who holds that key.

Managed · default

We run it for you.

Turn on encryption and your documents and search index are encrypted at rest under a key dedicated to your workspace. By default Contexel holds that key to run indexing and search. That is custodial: we can technically access your data, and we say so plainly.

Bring your own key · now live

Or hold the key yourself.

Switch to a key only you hold and supply it each session. It lives in your workspace's memory only while unlocked and is never written to disk; when the workspace is locked, Contexel has no way to decrypt your documents or your search index. Lose the key and the data is permanently unrecoverable, with no backdoor.

Enable it, rotate it, switch to your own key, or export your data, all from the console. Backups taken before you switch may retain a copy until they expire.

Straight talk

We tell you exactly what this is.

  • The hosted service is custodial today by default. A new workspace is not encrypted with a workspace key until you turn encryption on; when you do, by default we keep a copy of the key so we can run indexing and search for you, which means we can technically access your data. If you would rather we could not, bring-your-own-key is live now: switch to a key only you hold and supply it each session, and when the workspace is locked, Contexel has no way to decrypt your documents or your search index.
  • No proprietary format. Your content is plain Markdown, and the search index is built from it and can be rebuilt any time. There is nothing locked away to migrate and nothing to reverse-engineer. Export whenever you want.
  • Plain claims only. If we cannot back it up, we do not say it. No inflated security language, no marketing sleight of hand.

Get started

From sign-in to the right context in minutes.

1 · Sign in

Open the console and sign in with your browser using Google, GitHub, Microsoft, or SSO. No install. Your workspace is ready right away.

Open the console
2 · Connect your agent
claude mcp add --transport http contexel \
  https://contexel.ai/mcp \
  --header "Authorization: Bearer <YOUR_TOKEN>"

Mint a token in the "Connect your agent" panel and this line shows up filled in, ready for Claude Code, Cursor, or any MCP client. Where browser sign-in (OAuth 2.1) is turned on, add the address with no token and sign in from your editor.

Then just talk to it. Ask "what do we know about rate limiting?" and your agent calls get_context. Say "remember this for next time" and it calls observe, and a person approves anything lasting through review. Not sure what to say? There is a help tool for that.

Give your agent the right context.

Hosted for your team. Sign in and connect in minutes.

Open the console See the connect steps