Shared context for AI-forward teams
Every agent reads it.
Anyone can add to it.
Approve + review
according to you
reedright is your company's brain: one reviewed source of what you know. Claude, ChatGPT, and the agents you build read it before they answer. Your team browses it on a website. Underneath, it is plain files you own.
2 minutes to connect your first tool. No credit card. Nothing for your teammates to install.
reedright does one thing. It keeps the context your AI tools share correct, current, and yours.
1
Every tool reads the same thing.
Connect Claude, ChatGPT, Cursor, or an agent you built. Each person connects with their own login, so every read and every addition has a name on it.
Your agent reads the brain before it answers. When it learns something, it adds an entry.
| maria | Claude | read brand voice | 2 min ago |
| jules | ChatGPT | read refund policy | 14 min ago |
| ops-agent | agent | added an observation about returns | 1 h ago |
| dev | Cursor | read release checklist | today |
2
Anyone can add. The owner decides what stays.
Quick observations go in right away, with limits: short, no numbers, no duplicates. Rules and procedures wait for the owner of that topic.
The owner reads the change on one page and clicks approve or reject. No code, no engineers. The approval is written into the entry: who, when, and a link.
Each week, owners get a summary of what changed and what is waiting.
Say "membership", never "subscription". Members join; they do not subscribe.
Mention that a member can pause any month. Do not quote a price; link to the plans page instead.
- added by
- ops-agent
- source
- support calls, week 37
- added
- 12 Sep 2026
- review by
- 11 Mar 2027
3
Browse it like a website. Own it like a file.
Every brain gets a clickable, searchable website, free. Managers read what the agents know without a chat window. Search it. Click through it. Send a link to a colleague.
Pull in Google Docs. See who reads what and what is due for review. Underneath, it is plain text in a repository your company owns. Leave any time and keep everything.
- Refunds: what we promiserule · support
- Issue a refundprocedure · support
- Refund volume by monthref · finance
How it works
- 1
Sign up and connect a repository.
We lay it out for you.
- 2
Invite your team. Connect your tools.
One login per person. Nothing to install.
- 3
Agents read and add. Owners approve.
Everyone browses the same brain.
How teams use it
Ops lead, consumer brand
Every Monday I approve five or six things the agents noticed last week. Ten minutes. Before, nobody wrote them down.
Marketing manager
I have never opened a repository. I get a link, read the change, click approve. When I want to know what the agents think about our brand, I open the site and search.
Founder working with an agency
The agency's Claude and our ChatGPT finally agree on our brand voice, because they read the same entry.
Automation builder
My agent adds an observation after every run. The good ones get kept. The bad ones get rejected before anyone acts on them.
Questions
- What is shared context?
- The things your company knows that an AI tool needs before it answers: how you talk, what you decided, how you do things, where the numbers live. reedright keeps it in one place that every tool reads.
- Which AI tools work with it?
- Claude (claude.ai and Claude Code), ChatGPT, Cursor, and any agent that speaks MCP, the standard way tools connect to outside context. One connection per person. The tool reads and adds on your behalf.
- Do my teammates need GitHub?
- No. They sign up with an email. They approve from a web page. They browse the brain on its website. Only the first person connects the repository.
- What stops an agent from adding nonsense?
- Limits. Observations are short, cannot contain numbers, and are rejected if they duplicate an existing entry. Rules, procedures, and references wait for an owner. Every entry has an author, a source, and a review date. Anything can be rejected before it lands and removed after.
- Why not ChatGPT memory or a Claude project?
- Those belong to one person and one vendor. Nobody reviews them. They leave with the account. reedright is shared, reviewed, and yours.
- Why not a wiki?
- Agents cannot write to a wiki safely, and nobody keeps it current. In reedright, agents add, owners approve, and every entry has a review date. And you still get the website.
- Where is my data? What do you store?
- Your brain lives in your repository. reedright stores accounts, who added and approved what, and which entries were read. Never prompts. Never conversation content. Delete an entry by deleting the file. Uninstall and keep everything.
- What does it cost?
- Free for small teams, with no credit card. Larger teams will pay per member, and we will say so before anything changes.
- How long does setup take?
- About 2 minutes to sign up and connect your first tool. Inviting the team and pulling in documents takes as long as you want to spend.
- Can I leave?
- Any time. Uninstall reedright and the files stay in your repository, readable without us.
For engineers: every addition is a pull request opened by the reedright GitHub App on the person's behalf. Observations merge when lint passes; rules, procedures, and refs wait for a domain owner in OWNERS.yaml. The approval is stamped into the file's frontmatter. The server speaks MCP over Streamable HTTP with OAuth 2.1. The layout is the Shared Context Protocol, in every brain as SCHEMA.md. Apache 2.0. Read the README.
Give every AI tool the same brain.
Free for small teams. 2 minutes to connect the first tool.
Sign up free