Operator docs

Operate MCP servers with evidence.

MegaMCP is a product suite for teams building, monitoring, connecting, and selling MCP servers. Sentinel v0 is shipped. Foundry is in pilot. Gateway, Vault, Registry, and Inspector are planned.

Connect a clientView status
The MCP Ops platform.Current surface
LiveSentinel v0
PilotFoundry
PlannedGateway
PlannedVault

Suite status.

These docs are intentionally factual. They describe what is shipped, what is in pilot, and what is planned next.

Sentinel
Shipped v0

Reliability checks for MCP servers: reachability, initialize, tools/list, latency, run history, and schema drift.

Foundry
Pilot

A control plane for hosted builder installs, releases, billing posture, and human-approved agent operations.

Gateway
Planned

A single authenticated endpoint for teams that need to connect many MCP servers without scattering credentials.

Vault
Planned

Credential handling, HMAC patterns, and scoped secret access for sensitive MCP workflows.

Registry
Planned

Discovery built from real inventory, operator review, access terms, and Sentinel evidence where available.

Inspector
Planned

Interactive testing and debugging for MCP tools before they are wired into production agent workflows.

The operating pattern.

Every public claim should trace back to real inventory, checks, review notes, or access terms. That is how the platform earns trust without fake scores.

01Register the serverCapture the owner, transport, endpoint or command, auth posture, and lifecycle state.02Import real inventoryUse tools/list payloads as the source of truth. Do not publish invented tool catalogs.03Attach evidenceAdd Sentinel checks, security notes, access terms, and review notes as they become real.04Expose carefullyConnect clients only with scoped tokens, clear permissions, and an audit trail for risky actions.

Sentinel v0.

The first shipped product monitors the health of other MCP servers and gives operators a record they can inspect.

  • Check stdio and streamable HTTP targets for reachability and handshake health.
  • Verify initialize and tools/list so dead or malformed servers are caught early.
  • Track latency and run history for operator review.
  • Compare tools/list snapshots to detect schema drift before clients break.

Run the public CLI.

The npm package is the zero-signup path for checking an MCP server. Version 0.1.9 preserves JSON output artifacts for CI, release gates, and audit handoff, then turns saved Sentinel results into a customer-facing reliability audit draft.

npx --yes @megamcp/sentinel@latest check https://your-server.example/mcp --fail-on-degraded --latency-budget-ms 5000 --min-tools 3 --output sentinel-result.json
npx --yes @megamcp/sentinel@latest audit sentinel-result.json --output mcp-reliability-audit.md

Add --header "Authorization: Bearer ..." for protected MCP endpoints. The check command exits non-zero for down targets, fails degraded targets with --fail-on-degraded, and writes sentinel-result.json before exiting. The audit command reads that JSON and writes mcp-reliability-audit.md.

Connect a client.

Use this after you have an account and an API token from the dashboard. Keep bearer tokens out of commits and shared screenshots.

{
  "mcpServers": {
    "megamcp": {
      "url": "https://megamcp.com/mcp",
      "headers": { "Authorization": "Bearer mmk_your_token" }
    }
  }
}

Replace mmk_your_token with the token from your dashboard.

Safety standard.

Production MCP work should be boring in the right places: scoped access, confirmation for risky actions, reviewable evidence, and clear ownership.

Do publish

Real tools/list inventory, current access terms, security notes, Sentinel evidence when available, and a seller or operator contact path.

Do not publish

Fake reliability scores, invented usage numbers, vague security claims, or planned product behavior written as if it already shipped.

Need help?

For access, review, or operator questions, contact support@megamcp.com.

Request early accessSee pricing