Fairvalid Ai MCP Server

FairValid for AI assistants

Verify email addresses from your AI assistant, follow verification jobs, and read results in your FairValid workspace. This package connects to FairValid's hosted Model Context Protocol (MCP) server and includes a skill for handling verification, credits, retries, and result interpretation.

Server: https://mcp.fairvalid.com/mcp
Transport: Streamable HTTP
Publisher and support: FairValid · nayeem@fairvalid.com
Documentation: fairvalid.com/help/mcp

Connect your workspace

Use OAuth in a supported client. Sign in and approve access on FairValid, then select an existing Live API key covering the requested permissions. Its workspace, scopes, and lifetime spending cap apply to the connection. A workspace Owner or Manager can manage keys on the API keys page. Revoking the selected key ends its connections.

No client secret is bundled or required for automatic public client registration. Never paste a key, password, OAuth code, or token into chat. For a client that uses bearer headers, configure an existing key through its protected credential settings. A direct Test key can be used for sandbox fixtures; OAuth currently selects Live keys.

Supported package formats

Package configuration does not establish a marketplace listing or prove a successful connection in every client. Provider account access, permissions, review, and native client acceptance are separate steps.

For Claude Code 2.1.275 or later, install inside a session with /plugin install fairvalid --marketplace moonlitnayeem/fairvalid-ai. For older versions, use /plugin marketplace add moonlitnayeem/fairvalid-ai followed by /plugin install fairvalid@fairvalid. These forms follow the official installation reference. The marketplace installs the dedicated claude/ package. A maintainer can validate that folder with claude plugin validate /path/to/fairvalid-ai/claude --strict --profile directory. Users can review and load a local public checkout with claude --plugin-dir /path/to/fairvalid-ai/claude. For Gemini CLI, the public repository can be installed with gemini extensions install https://github.com/moonlitnayeem/fairvalid-ai. These commands require that the referenced checkout or published repository exists. Review client trust prompts and complete OAuth before using the tools.

What you can do

Examples of requests:

• “Show my FairValid credit balance.”
• “Show the current status and results for this FairValid job.”
• “Verify these email addresses with FairValid.”

The last request starts live work when connected through OAuth or a Live key. Only submit addresses you intend to verify and are authorized to process.

Credits and result interpretation

Live submissions can reserve credits and return an asynchronous job. Completed Valid and Invalid checks use one credit each. Genuine Unknown, unresolved catch-all, technical failures, and unprocessed items have no net verification charge. Cancellation does not reverse charges for completed checks; in-flight work may still finish. A spending cap does not add credits to a workspace.

An accepted job is not a completed result. Poll with backoff, honor retry hints, and follow the results cursor to read all pages. Keep pending and unprocessed items distinct from final results. Unknown or unresolved catch-all is not proof of validity; Valid is not a guarantee of inbox delivery.

Both submission tools require a stable idempotency_key. On an uncertain response, keep the same selected API-key identity, same key, and unchanged input. Reconnecting to a different API key can create another run even if the idempotency key is unchanged. Do not rotate keys or resubmit with a new key just to make an uncertain request appear successful.

Connection and acceptance tests

A connection or installation request does not authorize customer verification or job cancellation. For an authorized OAuth connection check, discover the six tools and call fairvalid_get_balance; that read spends no verification credits.

Default submission tests use a separate direct Test-key connection and the documented example.com fixtures. Sandbox work makes no live provider call and spends no verification credits; Test keys cannot read the live balance.

A controlled live acceptance test requires the workspace owner's explicit authorization. Stay within the authorized test addresses, workspace, selected Live key, and credit limit; live charges and idempotency safeguards still apply. Do not run live tests automatically, use customer lists as installation fixtures, or cancel customer jobs as part of installation tests.

The packages contain manifests, the workflow skill, examples, and brand assets. The optional Zed extension also downloads and runs a pinned local transport bridge and stores OAuth credentials in its extension work directory. No credentials are bundled. Client validation, live protocol checks, and native client tests must each be recorded according to what was actually performed; none implies a provider's approval or publication. See the static validation record.

Privacy policy

When you invoke a tool, the connected AI provider sends the supplied addresses, job identifiers, and other tool arguments to FairValid. FairValid processes them under its Privacy Policy and Terms. The AI provider also processes conversation and tool data under its own terms. Review those policies before connecting. The packages add no analytics. The Zed extension's local OAuth storage is described in its connection instructions. For support or data-handling questions, contact nayeem@fairvalid.com.

License and brand

The distribution code, manifests, and documentation are available under the MIT License. FairValid names and logos identify the publisher and do not grant trademark rights or permission to imply endorsement of another service. See brand guidance.