Discover MCP services
Search MCP presets and public directories, compare candidates, and assess endpoint health before connecting a server.
Find a Model Context Protocol (MCP) service that fits a workflow, then assess its transport, authentication needs, health, and quality before you mount it. Agent Studio combines curated presets with public and managed catalogs, so treat directory metadata as a starting point rather than a security guarantee.
Before you begin
You must be signed in to Agent Studio. Decide whether the runtime can launch a local stdio command or must reach a cloud SSE endpoint. Also list the minimum tools and data access your workflow needs.
An administrator must configure the provider credential before Agent Studio can return managed catalogs from Composio or Arcade. Smithery remote services can also require a Smithery key, while GitHub-backed lookups can be more constrained without a GitHub token. Never paste a provider API key into the directory search box.
Find a service
Open the MCP area from the Tool Registry, then select 1-CLICK ECOSYSTEM PRESETS. Search or choose a category to find a local or cloud template that can prefill the connection form.
Use a preset when its transport and capability match your job. A preset is a configuration shortcut; it does not remove the need to review the command, endpoint, and credentials before connecting.
Select BROWSE REGISTRY or PUBLIC REGISTRY (GLAMA & MCP.SO). Enter a capability or product name, such as issue tracking or Postgres, rather than a broad term such as tools.
The aggregated directory can include results from Glama, mcp.so, Smithery, Composio, Arcade, and other catalog feeds. A partial provider failure does not necessarily prevent other sources from returning results.
Apply source, category, transport, and language filters. Choose STDIO only when the Agent Studio host may execute the listed command. Choose SSE when the host can reach the remote endpoint and you can supply its required authorization.
Compare the owner, description, tags, repository link, and authentication indicator. A VERIFIED badge reports provider metadata; it is not an Agent Studio audit of the server or its publisher.
Open a candidate's details. Review ENDPOINT HEALTH for reachability and latency, and QUALITY SCORE for the available quality signals. For a repository-backed result, inspect its GitHub information, release activity, and README before deciding.
Health can change after the probe, and a reachable endpoint can still reject initialization or tool calls. Likewise, popularity, usage counts, or a quality score do not prove that a server is safe for confidential data.
For Composio, choose a toolkit or an explicit subset of tools. Agent Studio creates a managed MCP session; if no toolkit is specified, the provider request defaults to all toolkits. Prefer the smallest set that completes the workflow.
For Arcade, choose a named integration such as github or slack. Agent Studio reports whether its initialization probe succeeded, but it can still return the endpoint with verified: false when the probe fails. Investigate that state before mounting the integration.
Select 1-CLICK MOUNT or MOUNT only after reviewing the candidate. Confirm the prefilled name, transport, endpoint or command, and authentication requirements in the connection form.
When mounting completes, the result appears under CONFIGURED MCP SERVERS. A managed provider reports how many tools were discovered when it connects successfully.
Verify the result
Troubleshooting
Next, connect and test MCP tools before adding the service to a production workflow.