Add governed data access to the agent framework, backend service, or product workflow you already run. Use approved connections, run data work outside the prompt, and keep an audit trail as the work happens.
Use MarcoPolo from the agent framework, backend service, or product workflow you already run. The SDK gives your code the pieces agents need in production: live access, a managed place to execute data work, and governance that travels with each call.
Credentials resolved here · never in your code
Salesforcesales_read
Snowflakeanalytics_ro
Postgresapp_read
Jiraissues_readCall the live systems your tenant has approved: Salesforce, Snowflake, Postgres, HubSpot, files, and internal databases. Credentials stay in MarcoPolo.
Secure connections# The work runs here, not in the prompt.$ mp.run("slipped_deals.sql")→ salesforce_opportunities 18 rows→ jira_issues 64 rows✓ slipped_deals_joined 14 rows$ mp.preview("result#7c21")# Your code gets 14 rows · not 82.▍
Execute saved queries, scripts, and workspace commands. Keep large outputs as result handles, then preview or compose them when the agent needs the next step.
Governed workspaceWritten as your code runs · not reconstructed after
Every action keeps tenant, user, service account, connection, execution, and result context attached. The audit trail is created as the work runs.
Trust and governanceBuild with the data gateway as part of your application runtime. Your code chooses the connection, starts the work, handles the result, and keeps governance attached from the first call.
Your workflow can start from the connections already provisioned for a tenant, user, service account, or delegated actor.
Queries, scripts, shell commands, joins, previews, and exports happen in the gateway before the model sees the answer.
Identity, grants, effects, request IDs, executions, results, and errors stay attached to the SDK workflow.
Admins can manage setup in the web app. Assistants can use approved access through MCP. Your SDK workflow does not reimplement either one.
A tenant admin connects Salesforce and Snowflake. A LangGraph workflow uses the SDK to run approved operations, compose result handles, and return a clean answer to the application.
A SaaS product acts on behalf of an end user. The agent can use the tenant's approved connections, while every call still carries delegated identity.
An analyst asks a follow-up question in an MCP client after the product workflow runs. The same approved connections and saved operations are available there.
No. The SDK is for agents and agentic apps built in code. MCP is the assistant-facing surface over the same gateway.
Yes. The web app can own provisioning, consent, and sharing while the SDK consumes visible connections and approved operations.
Yes. SDK workflows and MCP calls create activity against the same identity, connection, execution, result, and audit model.
Same data gateway, same governance model. What changes is who is driving and where the work is written.
Provision access, review what agents did, and hold the controls in one place.
Read the Agent Console pageOne MCPOne governed tool surface that Claude, ChatGPT, Cursor, and Copilot can use.
Read the One MCP pageLiveConnect100+ live enterprise systems behind the same six verbs, scoped per user and agent.
Read the LiveConnect pageAdaptive ContextA knowledge map per connection: schema, business rules, and joins that already worked.
Read the Adaptive Context page
Embed governed connections, managed data execution, and auditability into the agents you ship.
