1
Quick scan
Connect external AI clients like ChatGPT and Claude to your workspace context securely using the Model Context Protocol (MCP).
MCP
Requests
Tasks
Approvals
2
Overview
MCP lets compatible external AI clients work with organization-scoped Flopinio context without turning those clients into uncontrolled writers. The production connector uses bearer-token authentication, policy-enforced reads, observable runtime actions, and staged approval-backed writes. Public ChatGPT App publication remains under review.
3
What MCP is for
Use MCP when you want an external AI client to read workspace context or stage supported actions against Flopinio records.
4
Current setup model
Workspace owners and admins with agent runtime access can issue MCP tokens from Settings. Add the Flopinio MCP server and bearer token to your compatible client configuration, then restart the client.
5
Write safety
MCP is a scoped integration for an authenticated Flopinio user; it is not a general autonomous agent or broad Slack replacement. MCP writes require manual review by default. A token explicitly permitted to self-approve may approve its own eligible writes for the same user and organization. Both paths keep a recorded approval and execution history, and required permissions still apply.
6
Next steps
Read MCP Connector.
Read Grounded Execution.
Common Questions
Is MCP access self-serve right now?
Yes, for workspace owners and admins with agent runtime access. MCP remains beta, and issued tokens can be revoked from Settings.
Can an external AI write directly into my workspace?
Supported writes follow the token's scoped permissions and approval mode. Manual review is the default; an explicitly permitted token may self-approve its own eligible writes for the same user and organization.
Open the related workspace surface
Use the linked product area after reviewing the page guidance and permission notes.