Topic · Model Context Protocol

MCP servers

A Model Context Protocol server gives an AI agent a declared, bounded set of tools instead of an open browser session. This domain runs one: a public, read only MCP endpoint that answers questions about the projects, stack, and profile published here.

Connect to the live endpoint

The endpoint is public and unauthenticated because everything it returns is already published on this site. Point any MCP capable client at it.

MCP endpoint

{
  "mcpServers": {
    "alexcinovoj": {
      "url": "https://alexcinovoj.dev/mcp"
    }
  }
}

Tools exposed

get_profile returns the canonical bio, role, and location.
list_projects returns the public project ledger with repository links.
list_mcp_servers returns the MCP servers running in production.
list_stack returns the tooling inventory behind the builds.
list_engagements returns the TechTide AI engagement types, without pricing.

How the surface is scoped

Every tool is declared in the manifest. There is no dynamic tool surface and no passthrough execution.
The endpoint is read only. No tool writes, mutates, or accepts a payload that reaches a data store.
Only content already published on this domain is exposed, so the endpoint cannot leak anything a crawler could not read.
Permission scoping follows the deny by default pattern documented in the AI production track.

MCP permission scoping pattern Agent governance

Projects behind it

ClawKeeper

Governed Claude agent control plane. A deterministic policy engine runs before every LLM call, autonomous spend defaults to zero, and the audit log is enforced by the database.

TechTide Harness Kit

Trust contracts for agent skills expressed as JSON Schema: risk tiers, tool permissions, deny-by-default egress, approval gates, and prompt-injection fixtures behind a wall of CI validation gates.