Connect Silverbucket to your AI client (MCP)

Connect Silverbucket to your AI client (MCP)

What is this?

If you use an AI assistant like Claude, Cursor, or Windsurf, you can connect it directly to your Silverbucket data using MCP (Model Context Protocol), an open standard supported by these tools. Once connected, you can ask your AI assistant questions about your Silverbucket data or have it make changes on your behalf, without switching apps or copy-pasting information.

Setup

    In Silverbucket, an admin creates an API token: Admin → System settings → Tokens tab → Create token.
    In your AI client's settings, add Silverbucket as an MCP server, using the token and your Silverbucket instance URL.
    Restart your AI client. Try asking it something like "List all active projects and who leads each one" or "Does Engineering have capacity next month?"

What you can do through it

Domain
Covers
Example prompt
customers_*
Client companies
"List all customers with active projects"
projects_*
Projects, budgets, leads, status
"Create a new project called Atlas for customer Northwind"
tasks_*
Time-tracking categories within projects
"Add a 'Code Review' task to Project Atlas"
actuals_*
Actual hours logged against tasks
"Delete the incorrect actuals entry from last Tuesday"
users_*
People: org node, capacity, employment dates
"List all Backend Developers in the Espoo site"
memberships_*
Resource slots on projects
"Add Jane Doe to Project Nova at 50% capacity"
allocations_*
Planned hours over time on a membership
"Show planned allocations for Jane Doe next month"
capacity_*
Team-level availability by org node/week
"What's Engineering's capacity next month?"
org_*
Organisational hierarchy
"Delete the deprecated 'Legacy Ops' org node"
roles_*
Role definitions
"Create a new role called Solutions Architect"
competencies_*
Skill definitions
"List all competencies under the DevOps category"
usergroups_*
Named user groups
"Create a user group for the Helsinki office"
dashboard_*
Read-only aggregate views
"Give me the project board data for Project Atlas"

Important: permissions and security

  • The AI client can do anything the token's owner could do in Silverbucket — nothing more, nothing less.
  • We recommend creating a dedicated user with limited permissions specifically for this integration, rather than using an admin account's token.
  • You can revoke access immediately at any time: Admin → System settings → Tokens tab → Delete token.
  • Tokens don't expire automatically, so revoking is a manual step when you no longer need the connection.

Which AI clients are supported

Claude (desktop app and web, signing in via SSO), Cursor, and Windsurf. If you use another AI client that supports MCP, check that client's own documentation. Its MCP compatibility is outside our control.
This feature is currently being rolled out gradually across customers, so it may not be available on your account yet.

FAQ

Is this safe to connect to our production data?

The connection uses the same permissions as whichever user created the token. We recommend a dedicated, limited-permission user for this rather than an admin account.

Can we revoke access if a token is compromised?

Yes, immediately, from the same admin screen used to create it.

Can we use this with ChatGPT or Gemini?

If your ChatGPT/Gemini client supports connecting to external MCP servers, yes in principle. Check your client's own MCP support, as we haven't verified every third-party client ourselves.