Quickstart: repo to MCP
Create and verify your first hosted MCP server from a GitHub repository or OpenAPI specification.
You need an Elva workspace with permission to create collections and MCP servers, an API source, and a reachable upstream API. Use a test account with read-only access for your first call.
Start here
Sign in at app.getelva.ai and select the workspace you want to use. You can explore the supplied demo and onboarding tour before connecting your own API.
Build your catalog
The living API catalog
Every API in one scored, searchable place, refreshed on each scan with owners and last commit.
Hosted MCP servers
Turn any collection into a hosted MCP server where you pick exactly which operations become tools.
Trust your contracts
API contracts
Compose a versioned, governed promise from your collections — field-level PII control, drift alerts, and one-click publishing included.
Spec quality & scoring
30+ checks across design, developer experience, AI-readiness, and security produce a grade you can trust.
Test and watch
API testing
Fire real requests through an SSRF-safe proxy and generate client snippets in seconds.
MCP insights & logs
See which agents call which tools, with latency, error rate, and token usage per server.
Go further
AI assistant & semantic search
Chat with your catalog and search specs semantically, with tool-calling over your collections.
The enterprise MCP gateway
Hosted tools use the configured upstream authentication and record call activity. Saved deployment credentials are encrypted at rest.
Follow the walkthrough below to reach a first verified tool call. Allow a repository scan to complete before reviewing its collection.
The path from repo to agent
Step | Action and success check |
1. Connect a source | Connect GitHub and choose a repository and branch, or import an OpenAPI file. Confirm you selected the intended workspace. |
2. Inspect the result | Wait for the scan or import to complete. Open the collection and confirm expected endpoints and schemas are present. |
3. Review quality | Inspect quality and coverage findings. Fix missing descriptions and schemas that affect the operations you want to expose. |
4. Configure an MCP | Create an MCP server from the collection. Name it, select a small read-only operation set, and configure the upstream base URL and authentication. |
5. Publish and connect | Publish the reviewed deployment. Open its install page, select your client, copy the generated setup instructions, and complete authentication. |
6. Verify a call | Ask for a known test record or list result. Confirm the response and find the call in MCP logs. Check the tool list contains only the intended operations. |
The upstream API must be reachable by the hosted service. A draft MCP is not callable. Collection refreshes do not automatically publish new tool definitions; review and publish changes separately.
Troubleshoot the first call
If no endpoints appear, inspect the source format or scan result. If authentication fails, verify scopes and reconnect. If the client sees old tools, confirm the deployment is published and reconnect the client. See Troubleshooting for scans, network access, authentication, and publishing failures.
On this page
- Quickstart: repo to MCP