Permission has a scope.
Choose no access, show in app, read, or read and write for each connection. Limit access further with database allowlists and result caps.
AI agents & MCP
A capable sidekick.
You hold the reins.
Connect an MCP-compatible coding agent to Bison’s database tools. Decide what it can see, where it can work, and which changes need your approval in the app.
Take a closer look Your agent
Bison MongoDB
Bison MongoDBDevelopment
Room to explore
Read & writeStaging
Look, don’t touch
Read onlyProduction
Off limits to agents
No access Connection permissions · Field redaction · Write approvals · Audit trail
Context on your terms.
Redact fields before data is returned to the agent. Show-in-app tools can display results in Bison without sending the documents back.
See what happened.
Review write requests in a native approval panel. Browse and export an audit log of agent requests when you want to retrace the work.
From first click to useful work
A place for
your next step.
- 01
Choose the connection’s policy
Enable the access level you need, then choose allowed databases and redaction rules.
- 02
Connect your MCP client
Use Bison’s client setup flow with tools such as Claude Code, Cursor, or Claude Desktop.
- 03
Work with a visible trail
Let the agent query or open results in Bison, review requested writes, and inspect activity in the audit log.
A few practical details
Good to know.
Before you begin.
Do I need AI to use Bison?
No. Agent access is optional. Browsing, queries, pipelines, dashboards, and exports all work independently.
Where does data shared with an agent go?
Bison returns only what the configured permissions allow. Data your agent receives is then subject to that agent and its model provider’s data policies.