MCP reference

The Veryfront MCP server lets assistants discover tools and act on Cloud resources or connected services. Use a trusted MCP client that supports Streamable HTTP.

Connect a client

SettingValue
Server URLhttps://api.veryfront.com/mcp
TransportStreamable HTTP
AuthenticationOAuth sign-in and consent. API keys remain an alternative

Connect an MCP client for setup and a first read-only call. A project-scoped connection uses /projects/{project_reference}/mcp on the same host.

Find and call a tool

After initialization, use your client's tool-discovery action to inspect available names and input schemas. A read-only call such as veryfront__list_projects verifies that the client can authenticate and retrieve accessible resources.

json
{
  "jsonrpc": "2.0",
  "id": 1,
  "method": "tools/call",
  "params": {
    "name": "veryfront__list_projects",
    "arguments": {"limit": 1}
  }
}

This is a request sent through an initialized MCP connection, not a complete standalone HTTP session. The tool reference provides arguments, access requirements, and an editable example for each callable tool.

Browse tools

WorkReference
Configure agent capabilitiesAgents
Manage runs, conversations, tasks, and triggersExecution
Run evaluations and inspect reportsEvaluations
Discover models and available AI toolsAI Gateway
Find and manage project knowledgeKnowledge
Work with connected servicesIntegrations
Manage projects, files, and branchesProjects
Manage releases and environmentsDeployment
Use isolated commands and filesSandbox
Reuse cached valuesCache
Investigate logs, metrics, and tracesObservability
Manage access and credentialsIdentity and Access
Inspect usage and spending controlsBilling and Usage

Understand availability and results

Discovery describes a tool's contract. It does not grant permission to call it. A tool can require project access, an external connection, or a particular execution context. Runtime-only tools are documented as agent capabilities and have no direct-call runner.

Inspect the JSON-RPC response and the tool result, including isError when present. Output schemas describe structured results where published. A tool without an output schema is marked explicitly. Do not assume its response has a fixed object shape.

The discovery tools expose catalog and error metadata. Tool names and compatibility explains canonical names and existing client behavior.

The platform server above executes tools. Use the API overview for shared concepts and the REST and GraphQL alternatives.