Official Platform MCP
Connect compatible AI tools to Globster through an official MCP interface.
Build with the Globster Plugin SDK, connect MCP servers, and share custom plugins through your own marketplace. Build around your tools, with scoped credentials and human approvals.
For an app already in our catalog, start with its managed connection. For a tool or process unique to your team, choose the extension that fits.
Explore integrationsBring an existing tool server into your agent’s work.
Use MCP when your internal service already exposes tools through the Model Context Protocol.
Choose a Streamable HTTP server or a stdio process supported by the agent’s runtime.
Configure the connection and credential references. Test tool discovery before enabling it.
Review the exposed tools, approve the connection changes, and enable the connection for the selected agent.
The MCP engine does not perform an OAuth login for arbitrary servers or guess legacy SSE endpoints. Managed app authorization is a separate connection flow. Availability depends on installed capabilities and network policy.
Use our Python SDK to turn your systems into agent capabilities.
Use a plugin when your integration needs its own tool definitions, logic, configuration, or lifecycle.
Build a Python plugin with the Globster Plugin SDK. Define focused tools with clear inputs, outputs, and expected side effects.
Declare your plugin’s version, SDK compatibility, and dependencies in plugin.toml. Register tools through the SDK and package your release with agent-plugin pack.
Test with a limited account and representative requests. Review credentials, approval behavior, and deployment compatibility before rollout.
The Plugin SDK exists today. Contact our team for SDK access and the supported development setup. Custom tools remain subject to the agent’s permissions and runtime policy.
Give your organization its own catalog of custom plugins.
Use a custom marketplace source to distribute the plugins and versions your team has reviewed.
Package and version your plugins. Publish a catalog with their descriptions, compatibility requirements, and release artifacts.
Connect your marketplace source to the intended agents. Keep internal packages on hosting with appropriate access controls.
Review installation and upgrade plans before rollout. Plugin compatibility and artifact integrity are checked during installation.
Custom marketplace sources are supported today. Private access requires suitable hosting and a permitted network path; contact us to agree the setup for your organization.
Turn team knowledge into instructions an agent can reuse.
Use a skill when the tools already exist and the agent needs a repeatable procedure or your team’s standards.
Give the skill a clear name and a description of when to use it.
Write concise instructions, the expected result, and the points that need human review.
Try the skill on a real example, review the output, and refine the instructions. The agent’s tool permissions still apply.
---
name: customer-brief
description: Prepare a customer brief before an account meeting
---
Summarize the account using its authorized sources.
Link the evidence and flag missing information.
Draft next steps for the account owner to review.Skills guide behavior; they do not grant access or execute code by themselves. The native harness supports SKILL.md files, and installed skill-management capabilities can manage user-authored skills.
More ways to build with Globster are on the way. These platform capabilities are planned and not yet available.
Connect compatible AI tools to Globster through an official MCP interface.
Integrate Globster into your own applications and services.
Work with Globster from your terminal and development workflows.
Register agents through a supported developer flow.
A few things to know before connecting your systems.
Still have a question? Talk to our team
Bring the system, the task, and your access requirements. We’ll help you choose the right integration path.
Review the security approach