Agents and API
Connect your agents
Add Noetix to Claude, ChatGPT, Grok Bot, Claude Code, Cursor, Codex, Windsurf or any MCP client, so your AI tools answer from your company's knowledge.
Noetix is an MCP server. Any agent that speaks MCP can connect to it and read your company's brain: search it, read its pages, follow your written procedures, and write back what it learns.
Every agent uses the same address:
https://mcp.noetix.dev/mcpHow signing in works
There is no key to copy. The first time an agent connects, it opens the Noetix sign-in in your browser and asks you to approve it. After that, the agent works as you: it sees your workspace's brain and nothing else.
Who you are comes from signing in, never from the address, so it's safe to share the address with your team.
To see the agents you approved, or to disconnect one, open Settings › MCP Server in Noetix. Settings › Agents marks each one you connected, and says whether your setup is up to date.
Set up your agent
Pick your agent. Your choice is remembered next time.
Claude
claude.ai, the desktop app and mobile. Add it once and it works in all of them.
Opens Claude with Noetix filled in. Check it and press Add.
https://mcp.noetix.dev/mcpOn a Team or Enterprise plan, an owner adds it for everyone.
Sign in. Press Connect and approve Noetix. Then, under Tool permissions, set Read-only tools to Always allow, so Claude doesn't ask before every search.
Set it up. Paste this into a new chat. It reads your brain and your profile, offers to bring in what Claude already remembers about you (saving only what you approve), and asks the brain a first question.
Set up Noetix in this chat app. Noetix setup for chat apps, version 1.
Noetix keeps two things apart, and you must never mix them:
- About me: my profile and what my assistants remember about me, private to me. Read it with `profile_read`, add to it with `memory_save`.
- My business: what my company knows and has decided, shared with my team if I have one. Read it with `brain_recall`; it lives on pages.
Never save a business fact as a memory about me, and never put something about me in the brain.
Do these steps in order, and keep each answer short.
1. Call `brain_status` and tell me in one line what my business brain holds and which sources are connected. If you don't have the Noetix tools, stop and tell me to add Noetix first at https://www.noetix.dev/settings/agents.
2. Call `profile_read` and tell me in two lines what Noetix knows about me.
3. Ask me: "Want me to bring in what I already remember about you?" If I say no, go to step 4. If I say yes, list what you remember about me that my profile doesn't already have: my preferences, my role and background, and what I'm working on now. Put each on its own line with where it would go (preference, about me, or current work). Save only the lines I approve, each with `memory_save`, and tell me what you saved. Never save anything about my company here.
4. Give me this text to paste where my chat app keeps instructions for every chat (Claude: Settings, Your Claude, Instructions. ChatGPT: Settings, Personalization, Custom instructions), because you can't change those settings yourself:
I use Noetix. At the start of a chat, call profile_read with setup "chat-1". For anything about my business, ask brain_recall first and cite what it returns. When we decide something or finish work, record it on the Notes page with page_edit. Save lasting preferences about me with memory_save, never business facts.
5. Ask my business brain one question with `brain_recall`, such as what the business does or a recent decision, and show me the answer with its sources. If it has nothing, say so and call `brain_flag_gap` with the question.
ChatGPT
Added on the web, on any plan. Then it works in every chat.
- In Plugins, press Add.
- Choose Add custom MCP server.
- Name it Noetix and paste the URL below.
- Tick I understand, then press Create.
https://mcp.noetix.dev/mcpSign in. Press Connect Noetix, then approve Noetix.
Set it up. Paste this into a new chat. It reads your brain and your profile, offers to bring in what ChatGPT already remembers about you (saving only what you approve), and asks the brain a first question.
Set up Noetix in this chat app. Noetix setup for chat apps, version 1.
Noetix keeps two things apart, and you must never mix them:
- About me: my profile and what my assistants remember about me, private to me. Read it with `profile_read`, add to it with `memory_save`.
- My business: what my company knows and has decided, shared with my team if I have one. Read it with `brain_recall`; it lives on pages.
Never save a business fact as a memory about me, and never put something about me in the brain.
Do these steps in order, and keep each answer short.
1. Call `brain_status` and tell me in one line what my business brain holds and which sources are connected. If you don't have the Noetix tools, stop and tell me to add Noetix first at https://www.noetix.dev/settings/agents.
2. Call `profile_read` and tell me in two lines what Noetix knows about me.
3. Ask me: "Want me to bring in what I already remember about you?" If I say no, go to step 4. If I say yes, list what you remember about me that my profile doesn't already have: my preferences, my role and background, and what I'm working on now. Put each on its own line with where it would go (preference, about me, or current work). Save only the lines I approve, each with `memory_save`, and tell me what you saved. Never save anything about my company here.
4. Give me this text to paste where my chat app keeps instructions for every chat (Claude: Settings, Your Claude, Instructions. ChatGPT: Settings, Personalization, Custom instructions), because you can't change those settings yourself:
I use Noetix. At the start of a chat, call profile_read with setup "chat-1". For anything about my business, ask brain_recall first and cite what it returns. When we decide something or finish work, record it on the Notes page with page_edit. Save lasting preferences about me with memory_save, never business facts.
5. Ask my business brain one question with `brain_recall`, such as what the business does or a recent decision, and show me the answer with its sources. If it has nothing, say so and call `brain_flag_gap` with the question.
Grok Bot
xAI's agents that do real work. Add Noetix once and every Bot can use it.
- In Grok Bot, open Connect apps.
- Choose Add a custom connector.
- Name it Noetix and paste the URL below.
https://mcp.noetix.dev/mcpOn a team, an admin may have to allow this address first.
Sign in. Sign in to Noetix in the window that opens, and approve it.
Set it up. Paste this into a new chat. It reads your brain and your profile, offers to bring in what Grok Bot already remembers about you (saving only what you approve), and asks the brain a first question.
Set up Noetix in this chat app. Noetix setup for chat apps, version 1.
Noetix keeps two things apart, and you must never mix them:
- About me: my profile and what my assistants remember about me, private to me. Read it with `profile_read`, add to it with `memory_save`.
- My business: what my company knows and has decided, shared with my team if I have one. Read it with `brain_recall`; it lives on pages.
Never save a business fact as a memory about me, and never put something about me in the brain.
Do these steps in order, and keep each answer short.
1. Call `brain_status` and tell me in one line what my business brain holds and which sources are connected. If you don't have the Noetix tools, stop and tell me to add Noetix first at https://www.noetix.dev/settings/agents.
2. Call `profile_read` and tell me in two lines what Noetix knows about me.
3. Ask me: "Want me to bring in what I already remember about you?" If I say no, go to step 4. If I say yes, list what you remember about me that my profile doesn't already have: my preferences, my role and background, and what I'm working on now. Put each on its own line with where it would go (preference, about me, or current work). Save only the lines I approve, each with `memory_save`, and tell me what you saved. Never save anything about my company here.
4. Give me this text to paste where my chat app keeps instructions for every chat (Claude: Settings, Your Claude, Instructions. ChatGPT: Settings, Personalization, Custom instructions), because you can't change those settings yourself:
I use Noetix. At the start of a chat, call profile_read with setup "chat-1". For anything about my business, ask brain_recall first and cite what it returns. When we decide something or finish work, record it on the Notes page with page_edit. Save lasting preferences about me with memory_save, never business facts.
5. Ask my business brain one question with `brain_recall`, such as what the business does or a recent decision, and show me the answer with its sources. If it has nothing, say so and call `brain_flag_gap` with the question.
Claude Code
Anthropic's terminal coding agent.
claude mcp add --transport http noetix https://mcp.noetix.dev/mcpSign in. In Claude Code, run /mcp, choose noetix and press Authenticate.
Set up your repo. In your repo, paste this into your agent. It adds Noetix to your CLAUDE.md and AGENTS.md, imports your team's skills, and asks the brain a first question. It never commits. Or run /mcp__noetix__setup.
Connect this repo to our company brain, Noetix. Noetix repo setup, version 1. Do these steps in order.
1. Call `brain_status` and tell me in one line what the brain holds and which sources are connected. If you don't have the Noetix tools, stop and tell me to connect Noetix first at https://www.noetix.dev/settings/agents.
2. Add the section below to this repo's agent instruction files, so every agent and every teammate who clones the repo knows about the brain:
- If the repo has both `CLAUDE.md` and `AGENTS.md`, add it to both, unless `CLAUDE.md` imports `AGENTS.md` (a line `@AGENTS.md`); then add it to `AGENTS.md` only.
- If it has only one of them, add it to that one.
- If it has a `.cursor` folder (check with `ls -a`, since an empty folder doesn't show in a file search), also write it to `.cursor/rules/noetix.mdc`, starting with this frontmatter:
```
---
description: Noetix, our company brain
alwaysApply: true
---
```
- If it has none of these, create `AGENTS.md` with the section.
- If a section starting `## Noetix: our company brain` is already there, replace it rather than adding a second one. Change nothing else in these files.
The section, exactly:
```markdown
## Noetix: our company brain
Noetix is our company brain, connected here as an MCP server. Every agent on the team
(Claude Code, Cursor, Codex and the Noetix assistant) reads and writes the same brain.
It knows what this repo can't tell you: what the company decided and why, how we do
things, what teammates' agents already did, and what was said in Slack, Notion, Drive,
GitHub and Linear.
- Before changing how something works, or when it isn't clear why code is the way it
is, ask `brain_recall`. Cite what it returns. If it has nothing, say so.
- Before a task the team may have a procedure for (a release, a review, a migration,
an incident), call `skill_search` and follow the match with `skill_use`.
- When you finish work a teammate would want to know about (a change merged, the cause
of a bug, a decision made with the person you're working with), record it with
`page_edit` on the page that covers it, or as a line on Notes:
`page_edit("Notes", "", "<what changed> (<PR, commit or file>)")`.
- Call `profile_read(setup="repo-1")` once at the start of a session: it holds what the
person you work for wants agents to know about them, private to them. When they
state a lasting preference or correct how you work with them, save it with
`memory_save`. Company facts never go there.
- Search the repo for code, not Noetix. Never write secrets, keys or customer data
to it.
```
3. Look for skills, commands and rules this repo's agents already use: `.claude/skills/*/SKILL.md`, `.claude/commands/*.md`, `.cursor/rules/*.mdc` (except `noetix.mdc`) and `.agents/skills/*/SKILL.md`. Show me the list, and import the ones I pick with `skill_import`, exactly as written. My personal skills in `~/.claude/skills` are often general tools rather than team procedures: just tell me how many there are, and list them only if I ask. If the repo has none, say so and go on.
4. Ask the brain one question about this repo with `brain_recall`, such as what it is for or a decision about it, and show me the answer with its sources. If the brain has nothing on it, say so and call `brain_flag_gap` with the question.
5. List every file you changed. Don't commit: I'll review and commit them myself, because they change how every teammate's agent works.
Codex CLI
OpenAI's coding agent.
[mcp_servers.noetix]
url = "https://mcp.noetix.dev/mcp"
auth = "oauth"Sign in. Run this, then approve in your browser.
codex mcp login noetixSet up your repo. In your repo, paste this into your agent. It adds Noetix to your CLAUDE.md and AGENTS.md, imports your team's skills, and asks the brain a first question. It never commits.
Connect this repo to our company brain, Noetix. Noetix repo setup, version 1. Do these steps in order.
1. Call `brain_status` and tell me in one line what the brain holds and which sources are connected. If you don't have the Noetix tools, stop and tell me to connect Noetix first at https://www.noetix.dev/settings/agents.
2. Add the section below to this repo's agent instruction files, so every agent and every teammate who clones the repo knows about the brain:
- If the repo has both `CLAUDE.md` and `AGENTS.md`, add it to both, unless `CLAUDE.md` imports `AGENTS.md` (a line `@AGENTS.md`); then add it to `AGENTS.md` only.
- If it has only one of them, add it to that one.
- If it has a `.cursor` folder (check with `ls -a`, since an empty folder doesn't show in a file search), also write it to `.cursor/rules/noetix.mdc`, starting with this frontmatter:
```
---
description: Noetix, our company brain
alwaysApply: true
---
```
- If it has none of these, create `AGENTS.md` with the section.
- If a section starting `## Noetix: our company brain` is already there, replace it rather than adding a second one. Change nothing else in these files.
The section, exactly:
```markdown
## Noetix: our company brain
Noetix is our company brain, connected here as an MCP server. Every agent on the team
(Claude Code, Cursor, Codex and the Noetix assistant) reads and writes the same brain.
It knows what this repo can't tell you: what the company decided and why, how we do
things, what teammates' agents already did, and what was said in Slack, Notion, Drive,
GitHub and Linear.
- Before changing how something works, or when it isn't clear why code is the way it
is, ask `brain_recall`. Cite what it returns. If it has nothing, say so.
- Before a task the team may have a procedure for (a release, a review, a migration,
an incident), call `skill_search` and follow the match with `skill_use`.
- When you finish work a teammate would want to know about (a change merged, the cause
of a bug, a decision made with the person you're working with), record it with
`page_edit` on the page that covers it, or as a line on Notes:
`page_edit("Notes", "", "<what changed> (<PR, commit or file>)")`.
- Call `profile_read(setup="repo-1")` once at the start of a session: it holds what the
person you work for wants agents to know about them, private to them. When they
state a lasting preference or correct how you work with them, save it with
`memory_save`. Company facts never go there.
- Search the repo for code, not Noetix. Never write secrets, keys or customer data
to it.
```
3. Look for skills, commands and rules this repo's agents already use: `.claude/skills/*/SKILL.md`, `.claude/commands/*.md`, `.cursor/rules/*.mdc` (except `noetix.mdc`) and `.agents/skills/*/SKILL.md`. Show me the list, and import the ones I pick with `skill_import`, exactly as written. My personal skills in `~/.claude/skills` are often general tools rather than team procedures: just tell me how many there are, and list them only if I ask. If the repo has none, say so and go on.
4. Ask the brain one question about this repo with `brain_recall`, such as what it is for or a decision about it, and show me the answer with its sources. If the brain has nothing on it, say so and call `brain_flag_gap` with the question.
5. List every file you changed. Don't commit: I'll review and commit them myself, because they change how every teammate's agent works.
Cursor
The editor. Global config, or per-project if you'd rather scope it.
{
"mcpServers": {
"noetix": {
"type": "http",
"url": "https://mcp.noetix.dev/mcp"
}
}
}Sign in. In Cursor Settings › Tools & MCP, press Connect next to noetix.
Set up your repo. In your repo, paste this into your agent. It adds Noetix to your CLAUDE.md and AGENTS.md, imports your team's skills, and asks the brain a first question. It never commits.
Connect this repo to our company brain, Noetix. Noetix repo setup, version 1. Do these steps in order.
1. Call `brain_status` and tell me in one line what the brain holds and which sources are connected. If you don't have the Noetix tools, stop and tell me to connect Noetix first at https://www.noetix.dev/settings/agents.
2. Add the section below to this repo's agent instruction files, so every agent and every teammate who clones the repo knows about the brain:
- If the repo has both `CLAUDE.md` and `AGENTS.md`, add it to both, unless `CLAUDE.md` imports `AGENTS.md` (a line `@AGENTS.md`); then add it to `AGENTS.md` only.
- If it has only one of them, add it to that one.
- If it has a `.cursor` folder (check with `ls -a`, since an empty folder doesn't show in a file search), also write it to `.cursor/rules/noetix.mdc`, starting with this frontmatter:
```
---
description: Noetix, our company brain
alwaysApply: true
---
```
- If it has none of these, create `AGENTS.md` with the section.
- If a section starting `## Noetix: our company brain` is already there, replace it rather than adding a second one. Change nothing else in these files.
The section, exactly:
```markdown
## Noetix: our company brain
Noetix is our company brain, connected here as an MCP server. Every agent on the team
(Claude Code, Cursor, Codex and the Noetix assistant) reads and writes the same brain.
It knows what this repo can't tell you: what the company decided and why, how we do
things, what teammates' agents already did, and what was said in Slack, Notion, Drive,
GitHub and Linear.
- Before changing how something works, or when it isn't clear why code is the way it
is, ask `brain_recall`. Cite what it returns. If it has nothing, say so.
- Before a task the team may have a procedure for (a release, a review, a migration,
an incident), call `skill_search` and follow the match with `skill_use`.
- When you finish work a teammate would want to know about (a change merged, the cause
of a bug, a decision made with the person you're working with), record it with
`page_edit` on the page that covers it, or as a line on Notes:
`page_edit("Notes", "", "<what changed> (<PR, commit or file>)")`.
- Call `profile_read(setup="repo-1")` once at the start of a session: it holds what the
person you work for wants agents to know about them, private to them. When they
state a lasting preference or correct how you work with them, save it with
`memory_save`. Company facts never go there.
- Search the repo for code, not Noetix. Never write secrets, keys or customer data
to it.
```
3. Look for skills, commands and rules this repo's agents already use: `.claude/skills/*/SKILL.md`, `.claude/commands/*.md`, `.cursor/rules/*.mdc` (except `noetix.mdc`) and `.agents/skills/*/SKILL.md`. Show me the list, and import the ones I pick with `skill_import`, exactly as written. My personal skills in `~/.claude/skills` are often general tools rather than team procedures: just tell me how many there are, and list them only if I ask. If the repo has none, say so and go on.
4. Ask the brain one question about this repo with `brain_recall`, such as what it is for or a decision about it, and show me the answer with its sources. If the brain has nothing on it, say so and call `brain_flag_gap` with the question.
5. List every file you changed. Don't commit: I'll review and commit them myself, because they change how every teammate's agent works.
Windsurf
Cascade reads its own config file, with its own key name.
{
"mcpServers": {
"noetix": {
"serverUrl": "https://mcp.noetix.dev/mcp"
}
}
}Sign in. In Cascade's MCP servers, press Authenticate next to noetix.
Set up your repo. In your repo, paste this into your agent. It adds Noetix to your CLAUDE.md and AGENTS.md, imports your team's skills, and asks the brain a first question. It never commits.
Connect this repo to our company brain, Noetix. Noetix repo setup, version 1. Do these steps in order.
1. Call `brain_status` and tell me in one line what the brain holds and which sources are connected. If you don't have the Noetix tools, stop and tell me to connect Noetix first at https://www.noetix.dev/settings/agents.
2. Add the section below to this repo's agent instruction files, so every agent and every teammate who clones the repo knows about the brain:
- If the repo has both `CLAUDE.md` and `AGENTS.md`, add it to both, unless `CLAUDE.md` imports `AGENTS.md` (a line `@AGENTS.md`); then add it to `AGENTS.md` only.
- If it has only one of them, add it to that one.
- If it has a `.cursor` folder (check with `ls -a`, since an empty folder doesn't show in a file search), also write it to `.cursor/rules/noetix.mdc`, starting with this frontmatter:
```
---
description: Noetix, our company brain
alwaysApply: true
---
```
- If it has none of these, create `AGENTS.md` with the section.
- If a section starting `## Noetix: our company brain` is already there, replace it rather than adding a second one. Change nothing else in these files.
The section, exactly:
```markdown
## Noetix: our company brain
Noetix is our company brain, connected here as an MCP server. Every agent on the team
(Claude Code, Cursor, Codex and the Noetix assistant) reads and writes the same brain.
It knows what this repo can't tell you: what the company decided and why, how we do
things, what teammates' agents already did, and what was said in Slack, Notion, Drive,
GitHub and Linear.
- Before changing how something works, or when it isn't clear why code is the way it
is, ask `brain_recall`. Cite what it returns. If it has nothing, say so.
- Before a task the team may have a procedure for (a release, a review, a migration,
an incident), call `skill_search` and follow the match with `skill_use`.
- When you finish work a teammate would want to know about (a change merged, the cause
of a bug, a decision made with the person you're working with), record it with
`page_edit` on the page that covers it, or as a line on Notes:
`page_edit("Notes", "", "<what changed> (<PR, commit or file>)")`.
- Call `profile_read(setup="repo-1")` once at the start of a session: it holds what the
person you work for wants agents to know about them, private to them. When they
state a lasting preference or correct how you work with them, save it with
`memory_save`. Company facts never go there.
- Search the repo for code, not Noetix. Never write secrets, keys or customer data
to it.
```
3. Look for skills, commands and rules this repo's agents already use: `.claude/skills/*/SKILL.md`, `.claude/commands/*.md`, `.cursor/rules/*.mdc` (except `noetix.mdc`) and `.agents/skills/*/SKILL.md`. Show me the list, and import the ones I pick with `skill_import`, exactly as written. My personal skills in `~/.claude/skills` are often general tools rather than team procedures: just tell me how many there are, and list them only if I ask. If the repo has none, say so and go on.
4. Ask the brain one question about this repo with `brain_recall`, such as what it is for or a decision about it, and show me the answer with its sources. If the brain has nothing on it, say so and call `brain_flag_gap` with the question.
5. List every file you changed. Don't commit: I'll review and commit them myself, because they change how every teammate's agent works.
OpenClaw
Self-hosted assistant gateway. Slack, Discord, WhatsApp, Telegram.
openclaw mcp add noetix \
--url https://mcp.noetix.dev/mcp \
--transport streamable-http \
--auth oauthSign in. Run this to sign in to Noetix.
openclaw mcp login noetixHermes Agent
Nous Research's headless runtime, usually on a server you own.
hermes mcp add noetix --url https://mcp.noetix.dev/mcp --auth oauthSign in. Run this to sign in to Noetix.
hermes mcp login noetixAny other MCP client
Cline, Zed, VS Code or your own. If it speaks MCP, it connects.
{
"mcpServers": {
"noetix": {
"type": "http",
"url": "https://mcp.noetix.dev/mcp"
}
}
}Sign in. Your client opens the Noetix sign-in the first time it connects.
Set up your repo. In your repo, paste this into your agent. It adds Noetix to your CLAUDE.md and AGENTS.md, imports your team's skills, and asks the brain a first question. It never commits.
Connect this repo to our company brain, Noetix. Noetix repo setup, version 1. Do these steps in order.
1. Call `brain_status` and tell me in one line what the brain holds and which sources are connected. If you don't have the Noetix tools, stop and tell me to connect Noetix first at https://www.noetix.dev/settings/agents.
2. Add the section below to this repo's agent instruction files, so every agent and every teammate who clones the repo knows about the brain:
- If the repo has both `CLAUDE.md` and `AGENTS.md`, add it to both, unless `CLAUDE.md` imports `AGENTS.md` (a line `@AGENTS.md`); then add it to `AGENTS.md` only.
- If it has only one of them, add it to that one.
- If it has a `.cursor` folder (check with `ls -a`, since an empty folder doesn't show in a file search), also write it to `.cursor/rules/noetix.mdc`, starting with this frontmatter:
```
---
description: Noetix, our company brain
alwaysApply: true
---
```
- If it has none of these, create `AGENTS.md` with the section.
- If a section starting `## Noetix: our company brain` is already there, replace it rather than adding a second one. Change nothing else in these files.
The section, exactly:
```markdown
## Noetix: our company brain
Noetix is our company brain, connected here as an MCP server. Every agent on the team
(Claude Code, Cursor, Codex and the Noetix assistant) reads and writes the same brain.
It knows what this repo can't tell you: what the company decided and why, how we do
things, what teammates' agents already did, and what was said in Slack, Notion, Drive,
GitHub and Linear.
- Before changing how something works, or when it isn't clear why code is the way it
is, ask `brain_recall`. Cite what it returns. If it has nothing, say so.
- Before a task the team may have a procedure for (a release, a review, a migration,
an incident), call `skill_search` and follow the match with `skill_use`.
- When you finish work a teammate would want to know about (a change merged, the cause
of a bug, a decision made with the person you're working with), record it with
`page_edit` on the page that covers it, or as a line on Notes:
`page_edit("Notes", "", "<what changed> (<PR, commit or file>)")`.
- Call `profile_read(setup="repo-1")` once at the start of a session: it holds what the
person you work for wants agents to know about them, private to them. When they
state a lasting preference or correct how you work with them, save it with
`memory_save`. Company facts never go there.
- Search the repo for code, not Noetix. Never write secrets, keys or customer data
to it.
```
3. Look for skills, commands and rules this repo's agents already use: `.claude/skills/*/SKILL.md`, `.claude/commands/*.md`, `.cursor/rules/*.mdc` (except `noetix.mdc`) and `.agents/skills/*/SKILL.md`. Show me the list, and import the ones I pick with `skill_import`, exactly as written. My personal skills in `~/.claude/skills` are often general tools rather than team procedures: just tell me how many there are, and list them only if I ask. If the repo has none, say so and go on.
4. Ask the brain one question about this repo with `brain_recall`, such as what it is for or a decision about it, and show me the answer with its sources. If the brain has nothing on it, say so and call `brain_flag_gap` with the question.
5. List every file you changed. Don't commit: I'll review and commit them myself, because they change how every teammate's agent works.
More about Claude
Add to Claude opens Claude with Noetix already filled in: check it, press Add, then Connect. Added once, it works on claude.ai, in the desktop app and on mobile. The short link noetix.dev/claude opens the same thing.
- On a Team or Enterprise plan, an owner adds it for everyone, and each person then presses Connect.
- Fewer questions. Claude asks before every tool by default, searches included. On the connector's page, under Tool permissions, set Read-only tools to Always allow, so it only asks before a write.
More about ChatGPT
On the web, open Plugins, press Add, choose Add custom MCP server, name it Noetix, paste the address above, tick I understand, press Create, then Connect Noetix. It works on every plan, the free one too.
More about Grok Bot
In Grok Bot, open Connect apps, choose Add a custom connector, name it Noetix and paste the address above. Sign in to Noetix in the window that opens and approve it. You can also ask a Bot to add the address for you.
- Every Bot can use it. Connectors belong to your Grok account, so you add Noetix once.
- On a team, an admin may have to allow the address first. Until then the connector shows as disabled by your team admin.
The setup prompt for chat apps
After adding Noetix to Claude, ChatGPT or Grok Bot, paste the setup prompt into a new chat. It:
- Keeps two things apart: what your assistant learns about you goes to your private profile, and what your business knows stays in the brain.
- Offers to bring in what your assistant already remembers about you, and saves only the lines you approve.
- Ends with a short text to paste into Claude's Settings › Your Claude › Instructions or ChatGPT's Settings › Personalization › Custom instructions, so every new chat uses Noetix.
Claude and ChatGPT each decide when to ask before a tool runs. Whatever they do, every tool runs with exactly your permissions in Noetix.