Iris MCP: use your approved answers and RFP data in Claude, ChatGPT, Copilot, Gemini and Cursor
Short answer: Iris MCP connects your Iris workspace to the AI assistant your team already uses: Claude, ChatGPT, Microsoft Copilot, Gemini or Cursor. Once an Iris admin turns it on, each person signs in with their Iris account and approves access. The assistant can then search your approved answers, ask Iris for sourced answers, check projects, assignments and pipeline, and, only if you grant write access, add content or update answers. MCP access is included in Iris pricing.
In Iris: an admin turns on Enable AI Assistant Connections (Settings → Security → Data and AI). Who this is for: teams already using an AI assistant at work who want it to answer from approved, governed content instead of guessing.
On this page
- What is Iris MCP?
- Which assistants work with Iris MCP?
- What can the assistant do with Iris?
- Example prompts
- How do you set it up?
- Security and permissions
- FAQ
What is Iris MCP?
MCP (Model Context Protocol) is an open standard that lets AI assistants call tools in other software. Iris runs an MCP server. When you connect it, your assistant can look things up in Iris and, if you allow it, make specific changes, all under your own Iris account.
The difference from pasting documents into a chat: the assistant reads your approved answers and current Knowledge Map through Iris, and Iris's answers come back with their sources.
Which assistants work with Iris MCP?
| Assistant | Connection |
|---|---|
| Claude | Iris MCP over OAuth |
| ChatGPT | Iris MCP over OAuth |
| Microsoft Copilot | Iris MCP over OAuth |
| Gemini | Iris MCP over OAuth |
| Cursor | Iris MCP over OAuth |
Iris supports OAuth with dynamic client registration, so assistants that support custom MCP connectors can register and connect without anyone handling an API key. Whether custom connectors are available in your assistant depends on that assistant's plan and admin settings.
What can the assistant do with Iris?
Iris MCP offers 15 read tools, an approved-answer export, and 5 write tools. Write tools need extra consent.
| Group | What the assistant can do | Access needed |
|---|---|---|
| Ask Iris | Ask a question and get a sourced answer from your Knowledge Map and approved answers | Read |
| Approved answers | Search your Approved Answers Library | Read |
| Projects | List projects, see project stats and details, read questions and answers | Read |
| Assignments | See what's assigned to you | Read |
| Win/loss | See win/loss stats (the Win/Loss feature is in beta) | Read |
| Pipeline | See your RFP pipeline | Read |
| Knowledge-base health | Check the health of your Knowledge Map content | Read |
| CRM opportunity | See the CRM opportunity linked to a project | Read |
| Export | Export approved answers to a file | Read |
| Add to Knowledge Map | Add a URL or a file | Write |
| Add an approved answer | Propose a new approved answer. It goes to the library inbox for review, not straight into the library. | Write |
| Update or approve project answers | Edit an answer's text, or approve it as yourself | Separate project-write consent |
Example prompts
Each prompt below maps to a tool in the table above. Company names are fictional.
| You ask | What Iris returns |
|---|---|
| "Search our approved answers for how we describe incident response." | Matching approved answers from your library |
| "Ask Iris: do we support SAML SSO, and with which identity providers?" | A sourced answer from your content |
| "What's assigned to me in Iris right now?" | Your open assignments, by project |
| "Show me the questions and current answers for the Northwind RFP." | The project's questions and answers you're allowed to see |
| "What's in our RFP pipeline?" | Your pipeline from Iris |
| "What's our win rate on finished projects this quarter?" | Win/loss stats (beta) |
| "Run a knowledge-base health check." | Knowledge Map health from Iris |
| "Which Salesforce opportunity is the Contoso questionnaire linked to?" | The linked CRM opportunity |
| "Export our approved answers tagged Security." | A file of those approved answers |
| "Add this URL to our Knowledge Map, tagged Product." (write) | The page is added to the Knowledge Map |
| "Save this answer as a new approved answer for 'Do you encrypt data at rest?'" (write) | A new entry in the library inbox, waiting for review |
| "Update my answer to question 14 in the Northwind RFP with this text, then approve it." (project write) | The answer is updated and approved under your name, through your normal workflow |
How do you set it up?
- Turn it on (admin, once). An Iris admin enables Enable AI Assistant Connections in Settings → Security → Data and AI.
- Add Iris to your assistant. In your assistant's connector or MCP settings, add Iris as a custom connector using the Iris MCP server address from your CSM.
- Sign in and approve. Your assistant opens an Iris sign-in and consent screen. Review what it's asking for and approve. Read access is the default; write access to the Knowledge Map and library, and write access to project answers, are separate choices.
- Ask. Start a chat and ask about your Iris content.
- Need more access later? Reconnect and approve the additional scope. Access is never upgraded silently.
Security and permissions
| Question | Answer |
|---|---|
| Who controls whether it's available? | An Iris admin. It's off until the org setting is turned on. |
| How does the assistant authenticate? | OAuth. Each person signs in to Iris and approves the connection. No shared passwords or API keys. |
| What can the assistant see? | Only what that person can see in Iris: the same role permissions and tag-based access apply. |
| How is scope narrowed? | Default tags per group or user pre-scope Ask Iris and MCP to the right product line or region. |
| Can the assistant change things? | Only with write consent. Adding content and editing or approving project answers are separate scopes. |
| Can it approve its own library answers? | No. Approved answers added through MCP go to the library inbox for review. |
| Is it audited? | Yes. Edits made through MCP are labelled "AI Assistant" in the answer's history, alongside who and when. |
| Can access be revoked? | Admins can see connected assistants and revoke a connection. |
| Where does the data go? | Results the assistant requests are returned into that assistant's conversation, under your agreement with that assistant's provider. |
| Is our data used to train models? | Iris never uses your data to train AI models. |
Try Iris on your own content
Try Iris free on an RFP you've already answered. Then connect your assistant and ask it about the result.
Iris is rated 4.8/5 on G2. MCP access is included in Iris pricing.
FAQ
What is the Iris MCP server?
It's the Model Context Protocol server Iris runs so AI assistants (Claude, ChatGPT, Microsoft Copilot, Gemini and Cursor) can search your approved answers, get sourced answers, and read project, assignment and pipeline data from Iris under your own account.
Does connecting ChatGPT mean Iris uses OpenAI to draft answers?
No. Iris's AI runs on Anthropic Claude models through AWS Bedrock; your content isn't sent to OpenAI for drafting. Connecting ChatGPT is your choice, and it sends the Iris results you ask for into your ChatGPT conversation.
Can the assistant edit or approve RFP answers?
Only if you grant the separate project-write permission when you connect. Edits go through your normal Iris rules and approval workflow and are labelled "AI Assistant" in the history.
Can an assistant add answers to our library?
It can propose them, with write consent. New answers go to the library inbox for a content manager to review.
Is Iris MCP an extra cost?
No. MCP access is included in Iris pricing, which is per active user.
Does our admin need to do anything?
Yes. An Iris admin must turn on AI assistant connections for your organization before anyone can connect.
See it live
Respond to RFPs
without the grind.
Book a demo and see your real auto-fill rate on a translation
RFP from your pipeline.
