Portkey MCP server

What the model can do

Portkey AI gateway: request, cost, token, latency, error, cache and feedback analytics over time, broken down by model, provider, workspace or a metadata key; gateway configs, prompts and their versions, prompt collections, workspaces and the audit log; when you ask, can record feedback on a trace, create, change, set the live version of or delete a prompt, create, change or delete a config, create a workspace and invite a person to the organisation.

Reads

• portkey_usagePortkey gateway analytics over time: one metric (requests, cost, tokens, latency with p50/p90/p99, errors, error rate, unique users, cache hit rate…
• portkey_usage_breakdownPortkey analytics for a period split by model, provider, workspace or the values of one metadata key: requests, cost and…
• portkey_configsGateway configs (routing, fallbacks, retries, caching): slug, name, status, workspace and dates; or pass slug for one config's JSON…
• portkey_promptsPrompt templates: id, slug, name, collection, model, status and dates, searchable by name and filtered by collection or workspace; or pass prompt_id…
• portkey_collectionsPrompt collections (folders): id, name, slug, workspace, parent, and how many prompts and sub-collections each holds.
• portkey_workspacesWorkspaces in the organisation (Admin key): id, slug, name, description, status, budget limits (cost or tokens, alert threshold…
• portkey_audit_logsThe organisation's audit log (Admin key, Enterprise plan): who changed what through the Portkey API or app, by time, method, path, resource type…

Acts, when you ask

These change something in Portkey. The model is told to use them only when you ask for exactly that, and each one is granted by name: per team on a shared connection, by the owner on a personal one.

• portkey_send_feedbackRecords feedback on the requests in a trace: a score from -10 to 10, an optional weight from 0 to 1 and metadata.
• portkey_create_promptCreates a prompt template in a collection, with its template text, model parameters and the provider (virtual key slug) it runs on.
• portkey_update_promptChanges a prompt.
• portkey_set_prompt_versionMakes a version of a prompt the default, so every caller of the prompt gets that version from now on.
• portkey_delete_promptPermanently deletes a prompt and its versions; anything still calling it by id or slug starts failing.
• portkey_create_configCreates a gateway config (routing strategy, targets by virtual key or provider slug, retries, caching) in a workspace.
• portkey_update_configChanges a gateway config by slug: its name, its JSON (replaced whole, saved as a new version) or its status.
• portkey_delete_configPermanently deletes a gateway config; requests that still name its slug start failing.
• portkey_create_workspaceCreates a workspace in the organisation (Admin key), optionally with a monthly or weekly budget in cost or tokens and an alert threshold.
• portkey_invite_userInvites a person by email to the Portkey organisation (Admin key) with an organisation role and a role in each workspace you name…

Connect it

First connect Portkey on Dashboard → Connections, then add Memtro to your client once. The Portkey tools appear alongside the tools of everything else you have connected; nothing is configured per connector.

claude plugin marketplace add memtro-org/claude-memtro
claude plugin install memtro@memtro --config api_key=sk-memtro-...

{
  "mcpServers": {
    "memtro": {
      "url": "https://platform.memtro.com/api/mcp",
      "headers": {
        "Authorization": "Bearer sk-memtro-..."
      }
    }
  }
}

{
  "mcp": {
    "memtro": {
      "type": "remote",
      "url": "https://platform.memtro.com/api/mcp",
      "headers": {
        "Authorization": "Bearer sk-memtro-..."
      }
    }
  }
}

Who may use it

The key identifies the person, and the server offers that person the tools they may use everywhere else in Memtro. A shared Portkey connection is used by the teams an administrator granted it to, tool by tool, on the Permissions page; a personal connection is used by its owner alone, and its tools that act stay off until the owner allows them. See Teams and permissions and Connectors.

Ask it like this

• › How much did we spend on gpt-4o last week?
• › Which model had the highest error rate this month?
• › What was our cache hit rate over the last 30 days?
• › Which prompts are in the support collection?
• › Who changed the production config this week?

Questions

Is this an official Portkey MCP server?

No. It is Memtro's MCP server with Memtro's Portkey connector behind it: the connector reads Portkey through its API with the credentials you connect on Dashboard → Connections, and Portkey publishes nothing here. The same endpoint carries every other system you connect and Memtro's memory tools.

What does it cost?

Nothing on top of Memtro. One price per organisation by its number of users, from $50 a month for up to 10 users after a 7-day free trial, with every connector, the MCP server, jobs and memory included. You bring your own model keys.

Can it change anything in Portkey?

Only when you ask, and only what it has been granted: record feedback (a score) on a request trace; create a prompt template; change a prompt, saving a new version; set which version of a prompt is live; delete a prompt; create a gateway config; change a gateway config that live traffic uses; delete a gateway config; create a workspace; invite a person to the organisation with a role. Every other tool reads, searches or lists. An administrator grants each acting tool by name per team, and a personal connection acts only after its owner allows it.

Which clients work?

Any MCP client that speaks Streamable HTTP: Claude Code, Cursor, Claude Desktop, OpenCode, Codex, Zed, Windsurf, Cline and the Anthropic and OpenAI SDKs. There is no local process to run; the server is remote and authenticated with your key.

Other MCP servers · Engineering and operations

• GitHub MCP server
• AWS CloudWatch MCP server
• Sentry MCP server
• ServiceNow MCP server
• Okta MCP server
• Twilio MCP server