// integrations
Your tools. Your data.
Now your agents can reach them.
Out of the box: Drive, OneDrive, Gmail, Outlook, Slack, Teams, Jira, Salesforce, HubSpot. With Custom Tools: any MCP server, any REST API, any webhook. Plus the Office add-in (Beta), and self-hosted OpenAI-compatible LLMs as just another provider.
// out of the box
Already connected. One click in admin.
OAuth flows already built. Per-user or workspace-wide. Shared across your team or scoped to a single zone.
Google Drive
OAuth · document library
OneDrive
OAuth · M365
Gmail
OAuth · read & draft
Outlook
Graph · M365
Slack
OAuth · channels + DM
Microsoft Teams
Graph · channels + meetings
Jira
OAuth · Atlassian
Salesforce
OAuth · object read/write
HubSpot
OAuth · CRM
All connections workspace-admin gated. Per-zone access. Disconnect any time.
MY CONNECTIONS · 4 / 11 ACTIVE
Your team inside Microsoft 365.
The In My Mind Office add-in puts your team inside the documents you already use. Task pane in Word, Excel, and Outlook. A Copilot agent that invokes your team from anywhere M365 Copilot runs.
Word
BETATask pane with tree zone picker, streaming chat, per-conversation model switch, Team/Solo toggle, insert-to-doc button. Draft inline. Send to a council. Pull from your library.
Excel
BETASame task pane. Push a selected range into a research zone, or have your team formula-author with explanations alongside. Excel formulas guidance live.
Outlook
BETAReading-pane task pane. Run a council against the current thread. Triage a long thread into a 3-bullet summary. Draft a reply with the right context already loaded.
M365 Copilot Agent
BETAInvoke "@In My Mind" from anywhere Microsoft Copilot runs, Teams, Word's Copilot pane, Outlook's Copilot composer. Reaches your team without leaving the doc.
In closed beta. Production packages built. Available for closed-beta workspaces, request access via your Sales contact or workspace admin. Contact Sales →
Agents with their own accounts.
Assign a Microsoft 365 identity to an agent. Its own mailbox. Its own calendar. Its own Teams presence. Mail it sends comes from that agent, signed and threaded under that identity. Workspace-admin gated. Scopable down to a single zone.
- →Per-agent M365 mailbox. "Maya, the Research Lead" has the inbox [email protected]. She reads, drafts, sends as Maya.
- →Per-agent calendar. Agent can be invited to meetings. Accept/decline. Take notes. Send the recap.
- →Zone-scoped by default. Maya's M365 identity only operates on email/calendar relevant to the Research zone. Other zones don't see her inbox, and she doesn't see theirs.
- →Identifiable, not impersonated. The signature line, the avatar, the "from" address are the agent's. Recipients see a clear AI identity, not a human one.
- →Audit-logged. Every mail sent, every meeting accepted, every calendar block created, recorded in the zone's audit trail. Workspace admin can see who-as-whom did what.
In beta. Configured via workspace admin → Agents → bind M365 identity. Customer's own M365 tenant; In My Mind requests only the scopes needed for the chosen zone.
WORKSPACE AGENTS · ZONE BINDING
Maya · Research Lead
Drafted 4 emails this week · attended 2 meetings
Daniel · Front Desk
Answered 23 calls today · sent 8 follow-ups
Elena · Sales Ops
Logged 47 calls · scheduled 6 demos this week
Why agent-owned identity?
An AI sending mail from a person's account is impersonation. An AI sending from its own clearly-labeled mailbox is delegation. Same outcome, defensible boundary.
Zone-level access scoping
Bind an agent's identity to a single zone. The agent's M365 access (mail, calendar, files) is filtered to what's relevant to that zone, keeps customer data out of internal research, internal R&D out of customer-facing agents.
Workspace-admin gated
Only workspace admins can bind a new identity. Per-zone, per-scope, per-agent grant. Granted scopes shown in admin UI and audit-logged on every change.
Same architecture for SSO
Where M365 doesn't fit, the same pattern applies to your SSO identity provider (Okta, Auth0, Google Workspace, custom). Agent IDs federate; access scopes follow.
// custom tools
Connect anything else.
Three ways to extend. No glue code. No proxy to host. Workspace admin pastes a URL or a spec. Your agents can call it in chat and on voice.
MCP
Model Context Protocol
Tap the growing ecosystem of AI tool servers. Point at an HTTPS Streamable-HTTP MCP server URL and its tools appear to your agents.
Hugging Face
huggingface.co/mcp
DeepWiki
mcp.deepwiki.com
Context7
mcp.context7.com
+ your internal servers, vertical specialists (bioinformatics, legal, finance)
OPENAPI / REST
OpenAPI 3.x spec import
Paste any REST API's OpenAPI spec. Every operation becomes a tool your agent can call. API-key-header or bearer-token auth.
Your CRM
Salesforce · HubSpot · Zoho · custom
Internal microservice
api.internal/v1/orders
Public data API
finance · weather · gov · logistics
Swagger 2.0 must be converted to OpenAPI 3.x first.
WEBHOOK
Outbound webhook
Wire an agent action to your own backend. File a ticket, push a record, kick off a workflow, notify a channel.
Zapier · n8n · Make
no-code automation
Custom backend
POST your-server.com/hook
Notification channels
Slack · email · SMS · Teams
// custom llm connections
Starts with frontier. Grows with you.
Today's leading frontier models. Tomorrow's. Specialized and custom models for the work that demands them. The catalog ships with the providers below, and any OpenAI-compatible endpoint connects as another first-class provider, self-hosted Llama, fine-tuned Mistral, a Together / Fireworks / Cerebras deployment, with its own per-zone routing rules.
Anthropic
Claude Opus · Sonnet · Haiku
OpenAI
GPT-5.5 · GPT-5.4 · GPT-4.1 · o-series
Gemini 3.1 Pro · 3.1 Flash · 2.5 Flash
xAI
Grok 4.3 · Grok Voice · Imagine
+ Your endpoint
any OpenAI-compatible
Self-hosted Llama / Mistral
Your fine-tuned domain model. Runs in your VPC. Paste base URL, In My Mind treats it as a first-class provider.
Paid OpenAI-compat services
Cerebras, Together AI, Fireworks, Anyscale, drop in their endpoint URL and your key. Mix freely with frontier providers on the same council.
Workspace-admin gated
Non-US-origin models behind explicit admin opt-in. Per-zone visibility. Cost caps and audit trails.
// in the wild
Real teams. Real connections.
// secure by default
Every connection, audit-controlled.
Every call is locked to public endpoints (no internal-network scanning). Time-boxed. Size-capped. Responses sanitized before the agent sees them. Anything that changes data asks for confirmation first, out loud on voice, with an approval modal on chat. Per-zone grants, give a tool to Sales but not HR. Security architecture →
Where Humans and AI collaborate.