Your data. Your agent. Your rules.
Connect your mail, meetings, and tasks to any agent you already use. Scoped, audited, and read-only by default. It can't be talked into sending by an email it just read.
ClaudeChatGPTCursor
ClineClaude Code
Mira (in-app)Every agent you connect is another copy of your data.
Point five agents at your accounts and you have granted five broad, standing permissions, with no shared record of who did what. One crafted email can tell any of them to forward your data.
- A separate, broad grant for every agent you add
- Standing write and delete access by default
- A prompt-injection email can tell a connected agent to forward secrets
- No shared record. Revoking means chasing grants across every tool
- One scoped key per person, read-only by default
- Send, delete, and external shares pause for an approval card
- Provenance gating: read an untrusted email and write tools switch off for that turn
- One audit log for the app and the agent. Revoke once, both stop
Up and running in three steps.
No platform migration, no standing access handed out. Connect what you already use and keep control of every scope.
Connect your sources
Link mail, meetings, and tasks in a few clicks, over OAuth or a scoped API key.
Set your scopes
Read-only by default. Grant send or delete only when you mean to, per person.
Bring any agent
Drop one scoped key into Claude, Cursor, or ChatGPT, or use Mira in-app.
Your inbox tool sees your inbox. Waydock sees your work.
Mail, meetings, tasks, and contacts in one context, about 47 tools, not a single silo. Your agent reasons across the Fireflies transcript, the Linear ticket, and the email thread together, because they finally sit in the same place.
- Mail and calendar from Gmail and Outlook
- Meetings and transcripts from Fathom and Fireflies
- Tasks and tickets from Jira and Linear
- Contacts and threads, linked across all of it
Built for the attack that already happened.
Provenance gating
Once a turn reads an untrusted email or transcript, send and delete are removed for the rest of that turn. A message in your inbox cannot tell your agent to act.
Per-user scoped keys
Least privilege by default. New keys are read-only. Full access is an explicit, danger-styled choice.
Approval cards
Send, delete, and external shares pause for a human click. Nothing leaves without you.
One audit log
The app and the agent write to the same log. Revoke a source once and both stop in the same moment.
Typed tools, not a chat box.
Generate a scoped key, drop it into Claude, Cursor, Cline, or ChatGPT, and your agent works from your real context through MCP. Typed tools with clear scopes, not a screen scrape, not a copy and paste.
Read the MCP docs ↗// claude_desktop_config.json { "mcpServers": { "waydock": { "transport": "streamable-http", "url": "https://waydock.ai/api/mcp/stream", "headers": { "Authorization": "Bearer wdmcp_••••" } } } }
Not technical? Mira does it in-app.
Most of your team will never set up an MCP client. Mira gives them the same secured context inside Waydock. Ask, and it works across your mail, meetings, and tasks, under the same scopes and the same audit log.
What should I prep before my 2pm with Acme?
call notesWAY-318latest threadYes, it drafts your replies and briefs your mornings.
Triage, drafted replies, and a morning brief come along once your agent has the context. They are outputs of the secure connection, not the reason it exists.
Can you confirm the new pricing and a date we can get this signed off internally?
Governance built in, not bolted on.
I can finally let my team point AI at our accounts without lying awake about it. Read-only by default, every action in one log, and I can pull a key in a single click.
Connect your work to your agent in under a minute.
Generate a scoped key, drop it into your MCP client, and your agent is working from your real context. Free while in beta. Disconnect any source whenever you want.