AI-Powered Faxing: The Complete Guide to Faxify MCP (2026)
When I first proposed adding an MCP server to Faxify, the initial reaction was understandable skepticism: who wants their AI assistant sending faxes? It sounded like solving a problem nobody had. The more we looked at it, the more we noticed that the workflows that still require fax are exactly the ones humans most want to automate away. Legal filings. Real-estate disclosures. These aren't going digital because the regulators they answer to haven't caught up. So instead of waiting for the world to abandon fax, we put it inside the AI tools developers and operators already use every day.
Worth being upfront about: Faxify MCP is early. We shipped it ahead of clear demand because Model Context Protocol, Anthropic's open standard for AI-tool integration, is going to reshape how AI agents touch external systems. Fax is a category where being early matters. The pieces of the AI ecosystem that need this aren't fully in place yet (ChatGPT Free still doesn't have native MCP, for one), but they're coming fast. The tool is live; production adoption is just starting. What follows is what we built and why we built it now.
This guide covers the full picture: what Faxify MCP is, how it works under the hood, what AI tools it connects to, the use cases we built it for, how it stacks up against the alternatives, and how to get started in about ten minutes.
What Is Model Context Protocol?
The Model Context Protocol (MCP) is an open standard, originally introduced by Anthropic in November 2024, for connecting AI assistants to external systems — databases, APIs, file systems, internal tools — through a uniform interface. Before MCP, every AI tool integration was bespoke: Cursor had its own extension system, Claude Desktop had a different one, ChatGPT had plugins (later replaced by GPTs, then by the developer-mode MCP support), and so on. Building an integration once didn't mean it worked anywhere else.
MCP changes that by defining a protocol. An MCP server exposes a typed set of "tools" (function-like operations the AI can call), "resources" (data the AI can read), and "prompts" (parameterized templates). The AI client speaks MCP, the server speaks MCP, and any tool you build once works everywhere MCP is supported. Think of it as the "REST API" of the AI-tool integration era, except designed for the conversational, agentic patterns LLMs actually use, with streaming, structured outputs, and stateful interactions baked in.
By 2026, MCP is supported natively or via developer mode in every major AI surface: ChatGPT, Claude, Cursor, Continue.dev, Cline, and any OAuth 2.1-capable assistant. That's the surface area we built Faxify MCP to cover.
What Is Faxify MCP?
Faxify MCP is a hosted Model Context Protocol server that turns any MCP-compatible AI assistant into a fax client. It exposes 18 typed tools for sending, receiving, drafting, searching, and managing faxes, all through the same Faxify account you use in the iOS, Android, and web apps. No separate signup, no separate billing, no separate page allowance.
The hosted part matters. You don't run an MCP server yourself, and you don't deploy infrastructure. You point your AI assistant at https://mcp.faxify.com/api/v1/mcp, complete an authentication flow (more on that below), and the tools appear in your assistant's tool palette. From there, you can fax in natural language: "send the signed agreement at this path to 555-123-4567 with my cover page." The assistant calls send_fax, Faxify validates and dispatches, and the status flows back into the conversation.
Faxify MCP is built for three kinds of users. Individuals who'd rather not switch apps when they need to fax something mid-conversation in ChatGPT or Claude. Operators who want to automate recurring fax workflows: invoice acknowledgments, vendor onboarding forms, signed releases. Developers building dedicated agents that combine Faxify with other MCP servers like filesystem, email, or CRM to create end-to-end document workflows. These are the workflows we built the tool for; production adoption is still in its early stages.
You can explore the server directly at mcp.faxify.com.
How Faxify MCP Works Under the Hood
Faxify MCP is a single HTTP endpoint at https://mcp.faxify.com/api/v1/mcp that speaks the Model Context Protocol over JSON-RPC. Different AI clients reach it two different ways, depending on whether they speak OAuth natively.
Claude Web (OAuth) Cursor / Claude Desktop ChatGPT (dev mode) Continue.dev / Cline (OAuth 2.1 native) (stdio-based) │ │ │ OAuth 2.1 DCR │ stdio │ ▼ │ ┌──────────────────────────┐ │ │ @faxify/mcp-client │ │ │ (npm package) │ │ └──────────────────────────┘ │ │ │ │ HTTPS + JWT ▼ ▼ ┌────────────────────────────────────────────────────────┐ │ Faxify MCP Server │ │ https://mcp.faxify.com/api/v1/mcp │ └────────────────────────────────────────────────────────┘ │ │ JSON-RPC tools ▼ ┌────────────────────────────────────────────────────────┐ │ Faxify Backend │ └────────────────────────────────────────────────────────┘ │ │ fax transmission ▼ ┌────────────────────────────────────────────────────────┐ │ SignalWire carrier network │ └────────────────────────────────────────────────────────┘ │ │ T.38 / G.711 ▼ Recipient fax
The two authentication flows
OAuth 2.1 with Dynamic Client Registration (RFC 7591), for HTTP-native clients. ChatGPT Developer Mode and OAuth-capable assistants discover Faxify's OAuth endpoints automatically via the well-known metadata at /.well-known/oauth-authorization-server (RFC 8414). The client self-registers without needing a whitelist, completes the standard authorization-code flow against your Faxify account (Google SSO, Apple SSO, Email/OTP), and receives a stateless JWT authorization code encrypted with AES-256-GCM. No client secret to leak, no redirect URI whitelist to maintain.
JWT tokens, for stdio-based clients (Cursor, Claude Desktop, Continue.dev, Cline). You sign in at mcp.faxify.com, copy a JWT from Settings → MCP Configuration (the same page shows when it expires), and paste it into your client's MCP config as the FAXIFY_JWT_TOKEN environment variable. The @faxify/mcp-client npm package handles the stdio-to-HTTP bridge so your local AI tool can talk to the hosted MCP server over a child-process pipe.
The 18 typed tools
| Category | Tools |
|---|---|
| Drafts & uploads | create_draft, get_draft, update_draft, delete_draft, list_drafts, get_draft_preview, request_upload_url |
| Send | send_fax, send_fax_using_draft, resend_fax |
| Inbox & status | list_faxes, get_fax, get_fax_status, delete_fax |
| Account & plans | get_profile, get_subscription, list_plans, list_offerings |
Each tool is typed end-to-end. Input parameters, success responses, and error shapes are declared in the MCP schema, so your AI assistant gets autocomplete-quality argument suggestions and validates inputs before calling. The draft tools matter more than they look: an AI agent that drafts a fax, previews it, and waits for the user to confirm before calling send_fax_using_draft is the difference between "useful assistant" and "loose cannon."
What happens after send_fax
Faxify accepts the request, renders the document at the page-quality tier defined by your plan (Standard / Fine / Superfine), and dispatches the transmission through SignalWire's programmable fax infrastructure. SignalWire handles the T.38 (or G.711 fallback) negotiation with the recipient's fax endpoint, retries transient failures, and reports delivery status back to Faxify. The AI client polls get_fax_status or subscribes to webhook notifications until the fax reaches a terminal state (delivered, failed, busy, etc.).
This architecture is the actual differentiator. We didn't build a fax stack from scratch. We put a well-typed MCP layer on top of a carrier-grade fax provider that's been operating at scale for years. SignalWire's carrier network does what carrier providers do; Faxify MCP does what AI providers don't yet do.
The Provider Chain
⚠️ Compliance note: Faxify is not HIPAA compliant and does not sign Business Associate Agreements. Do not send PHI through Faxify. Faxify's carrier and document-storage providers hold their own BAAs with Faxify; that covers their relationship with us and extends no HIPAA coverage to you.
Here's an overview of who handles what:
| Layer | Provider | Notes |
|---|---|---|
| Fax transmission (T.38, G.711) | SignalWire | |
| Document storage | Google Cloud Storage | |
| Database (metadata, account records) | Neon PostgreSQL (Vercel) | |
| Authentication | Firebase Authentication | |
| Website, web app, API, and MCP server hosting | Vercel | |
| Document conversion | DigitalOcean | |
| Product analytics (aggregate behavior signals) | Amplitude | Tracks aggregate product usage — page views, feature events — not fax document contents |
| Faxify (the software) | Const Agility, LLC | Not HIPAA compliant; signs no BAA. Routes through the chain above |
Faxify's carrier and document-storage providers hold their own BAAs with Faxify; that covers their relationship with us and extends no HIPAA coverage to you. Do not send PHI through Faxify. Product analytics via Amplitude sees usage signals, not document contents.
Data Deletion
- Per-fax / per-draft deletion. When a user deletes a fax or draft — from the iOS app, Android app, web app, or via MCP using
delete_fax/delete_draft— the fax document is cleared from our document storage (Google Cloud Storage) and from carrier records at SignalWire. Minimal metadata required for billing, legal, tax, and audit purposes is retained per the published Privacy Policy. - Account deletion. Users can delete their entire account directly from the Faxify iOS app, Android app, or web app (app.faxify.com) — no email or support ticket required. Email-based deletion via hello@faxify.com is available as an alternative for users who can't access the apps.
- Automatic retention limits. On free accounts, faxes and drafts are kept for one month: faxes sent or received more than one month ago, and drafts last edited more than one month ago, are deleted automatically, together with their files and transmission metadata. When a paid subscription ends, the same rule applies after a 30-day grace period. See the Privacy Policy and Terms for canonical details.
For workflows that need shorter deletion windows or stronger retention controls than these standard limits, reach out at hello@faxify.com.
Supported AI Tools
Faxify MCP works with two categories of AI client. Both reach the same server at https://mcp.faxify.com/api/v1/mcp; only the transport and the authentication mechanism differ.
Native OAuth 2.1 clients
Use this path when your AI client supports native MCP-over-OAuth (no wrapper needed):
- Claude (Desktop with web auth, claude.ai web) — OAuth-native MCP. Add the server URL, complete the OAuth flow, done.
- ChatGPT (Developer Mode) — OAuth-native MCP works on the Plus, Pro, Business, Enterprise, and Education tiers, where Developer Mode is available on the web. Turn it on in Settings → Security and login, create a developer-mode app for
https://mcp.faxify.com/api/v1/mcpin ChatGPT Plugins (chatgpt.com/plugins, + button), and authorize. ChatGPT connects directly over OAuth and can't use the stdio bridge; the Free tier has no Developer Mode (see below).
Stdio-based clients (via @faxify/mcp-client + JWT)
This is the universal fallback. It works on every AI client that supports MCP-over-stdio. You generate a JWT in your Faxify account settings, drop it into the client's config file, and the @faxify/mcp-client npm package handles the stdio-to-HTTP bridge.
- Cursor —
~/.cursor/mcp.json. - Claude Desktop —
~/Library/Application Support/Claude/claude_desktop_config.json(macOS) or%APPDATA%/Claude/claude_desktop_config.json(Windows). - Continue.dev, Cline, and any other stdio-based MCP client — same pattern.
Example configuration for stdio clients:
{ "mcpServers": { "Faxify": { "command": "npx", "args": ["-y", "@faxify/mcp-client"], "env": { "FAXIFY_MCP_URL": "https://mcp.faxify.com/api/v1/mcp", "FAXIFY_JWT_TOKEN": "YOUR_JWT_TOKEN_HERE" } } } }
The wrapper is open source under the MIT license — the Faxify/mcp-client repository on GitHub has the full implementation of the @faxify/mcp-client npm package.
What about ChatGPT Free?
Honest answer: there's no MCP path on ChatGPT Free today. Free accounts don't have Developer Mode, and ChatGPT can't use the stdio bridge. For now:
Wait for a ChatGPT Apps directory listing — getting Faxify into the ChatGPT Apps directory is on our roadmap. Once it ships, Plus and Free users would get a one-click connect without needing to deal with config files. No firm timeline yet.
Five Use Cases We Built For
Five categories where fax still matters and where AI agents have a real role. We don't have a corpus of customer case studies yet — Faxify MCP is new enough that production adoption is just beginning. What follows is the workflows we built the tool to support and where we expect the strongest fit.
1. Procurement and vendor onboarding
An office manager pastes a vendor onboarding form into Claude, asks it to extract the relevant fields and draft a formatted fax to the supplier's fax line, then confirms and sends — all without leaving the conversation. The draft → preview → send pattern (using create_draft, get_draft_preview, then send_fax_using_draft) is the right shape for any workflow that needs a human sign-off: the AI never auto-sends; a human always reviews the preview first.
2. Legal case filings and service of process
Solo practitioners and small firms still file motions, serve subpoenas, and exchange discovery materials over fax, especially in jurisdictions where the court accepts fax filings and email isn't recognized for service. With Faxify MCP, an attorney can ask their AI assistant to format a motion, generate the cover sheet with case caption and certificate of service, and dispatch to the clerk's fax line — without copying files into a separate fax app. The same get_fax_status tool that lets the AI confirm delivery also gives the firm a defensible record of when the transmission completed.
3. Real-estate disclosures and signature workflows
A closing coordinator drafts seller disclosures and FHA addenda in their CRM, asks an AI agent to package them as a fax to the cooperating broker's office, and uses the inbox tools (list_faxes, get_fax) to track incoming signed copies. For many older brokerages and title agencies, fax remains the preferred channel because of how the audit trail works — and an AI that can both send outgoing disclosures and ingest incoming signed pages turns what was a multi-day phone-tag exercise into a same-day workflow.
4. Accounting and finance: invoice and statement automation
Accounts-payable teams at clients who insist on faxed invoices (more common than you'd think in industrial supply and government contracting) can now have their accounting system pipe invoice PDFs into an AI agent that sends, polls for delivery, and writes the confirmation back into the ERP. The send_fax, get_fax_status, and list_faxes tools cover the full lifecycle.
5. Recruitment and HR: regulated-industry hiring
Background-check authorizations, employment-verification releases, and offer letters in regulated industries (defense contracting, financial services) sometimes still require fax-based delivery to upstream verification providers. An AI agent that drafts the document, attaches the appropriate authorization form, and faxes to the verification service — then files the confirmation back to the candidate record — replaces the fax-machine-in-the-corner-of-HR workflow that nobody has time for anymore.
Faxify MCP vs the Alternatives
If you're evaluating Faxify MCP, you're probably comparing it against either rolling your own integration on top of a carrier API, or using a programmable fax API directly without the MCP layer.
| Faxify MCP | Telnyx Programmable Fax | SignalWire Fax API | Build your own MCP | |
|---|---|---|---|---|
| Purpose | AI-native fax for any MCP client | Carrier-grade fax HTTP API | Carrier-grade fax HTTP API | DIY on top of any provider |
| MCP server included | ✅ Hosted | ❌ Build yourself | ❌ Build yourself | You're building it |
| OAuth 2.1 with DCR | ✅ | ❌ | ❌ | DIY |
| 18 typed tools | ✅ Pre-built | None | None | DIY (~2 months of work) |
| Free tier | 5 pages/month (no card) | Pay-per-use only | Pay-per-use only | Plus your provider cost |
| Paid plans | $7.99–$14.99/month (from ~$5.83/mo on annual) | ~$0.02/page no included pages | Variable | + dev time |
| HIPAA infrastructure | Not HIPAA compliant; signs no BAA | Via Telnyx (BAA) | Via SignalWire (BAA) | Whatever you wire up |
| Best for | Individuals, operators, and devs who want AI fax now | Engineering teams building bespoke fax infrastructure | Same as Telnyx | Teams with very specific custom needs |
Be honest about when Faxify MCP isn't the right answer: if you're building a fax product that competes with us, Telnyx or SignalWire directly is the better starting point. If you're embedding fax into your own SaaS at high volume with custom routing logic, the same applies. Faxify MCP is for users and operators who want to use fax inside their AI tools, not for teams who want to rebuild fax from the carrier API up.
For the much larger group of users who fall into the first category — anyone whose primary job isn't fax infrastructure — building your own MCP server on top of a carrier like Telnyx is a roughly two-month engineering project to deliver something that does less than @faxify/mcp-client does today.
Pricing and Page Limits
Faxify MCP is included in every Faxify plan, including the free tier. There is no separate MCP subscription and no surcharge for AI-initiated faxes versus app-initiated ones. MCP usage decrements the same monthly page allowance you'd use from the iOS, Android, or web apps.
- Free: 5 pages/month, no credit card required (includes ads in the apps; MCP usage has no ads).
- Basic (closed to new sign-ups): $4.99/month or $39.99/year — 100 pages/month, dedicated fax number on annual plans, Standard page quality.
- Standard: $7.99/month or $69.99/year — 250 pages/month, dedicated fax number included (monthly and annual), Standard quality.
- Professional: $9.99/month or $89.99/year — 750 pages/month, Fine quality.
- Elite: $14.99/month or $129.99/year — 2,000 pages/month, Superfine quality.
The free tier is genuinely useful for evaluating Faxify MCP — 5 pages is enough to wire up your AI assistant, run a half-dozen real test sends, and decide whether the integration earns its keep before you put a card down. Current pricing on the pricing page is canonical.
For higher monthly page volumes than the published plans (more than 2,000 pages/month), unlimited usage tiers, or any other special requirements, contact us at hello@faxify.com or +1 (945) 398-0695.
Faxify operates in the US and Canada.
Getting Started in 10 Minutes
The fastest path depends on your AI client.
For Claude Desktop — Open ~/Library/Application Support/Claude/claude_desktop_config.json (or the Windows equivalent), add the JSON snippet shown earlier with your JWT token (copied from Settings → MCP Configuration at mcp.faxify.com), and restart Claude Desktop. Verify the connection by asking Claude "list the Faxify tools available" — you should see all 18.
For Cursor, Continue.dev, or Cline — Same shape, in each client's MCP config file (~/.cursor/mcp.json for Cursor, the equivalent for the others). Same JWT environment variable.
For ChatGPT Plus, Pro, Business, Enterprise, or Education — On the web, enable Developer Mode in Settings → Security and login, open ChatGPT Plugins (chatgpt.com/plugins), click + to create a developer-mode app for https://mcp.faxify.com/api/v1/mcp, and complete the OAuth flow. The 18 Faxify tools appear in your tool palette once authorization is granted.
For ChatGPT Free users — Native MCP isn't available on this tier (no Developer Mode), and ChatGPT can't use the stdio bridge. For now, wait for the ChatGPT Apps directory listing (in progress).
Once connected, try this as your first prompt: "Show me my Faxify subscription, the last 5 faxes I sent, and how many pages I have left this month." This single conversation will exercise get_subscription, list_faxes, and get_profile, and verifies your auth is working end-to-end before you send anything.
For deeper technical detail — full auth flow walkthroughs, error response formats, and per-client configuration — the documentation lives at mcp.faxify.com/en/docs.
FAQ
What is the Faxify MCP server URL?
The Faxify MCP server endpoint is https://mcp.faxify.com/api/v1/mcp. Stdio clients (Cursor, Claude Desktop) reach it through the @faxify/mcp-client npm wrapper with a JWT token. Direct HTTP clients (ChatGPT Developer Mode, custom MCP integrations) connect to the same URL using OAuth 2.1.
Does Faxify MCP cost extra, or is it included in my Faxify plan? Faxify MCP is included in every Faxify plan, including the free tier. MCP usage decrements the same monthly page allowance as the regular Faxify apps — 5 pages/month on free, 100 to 2,000 pages/month on paid plans. There is no separate MCP subscription.
Which AI assistants work with Faxify MCP?
Faxify MCP works with stdio-based AI clients (Cursor, Claude Desktop, Continue.dev, Cline, and any other stdio MCP client) through the @faxify/mcp-client npm wrapper using a JWT token. Native OAuth-MCP connections work in Claude (Desktop and claude.ai web) and in ChatGPT Plus, Pro, Business, Enterprise, and Education (Developer Mode, on the web). ChatGPT Free has no Developer Mode and can't use the stdio bridge — those users can wait for the ChatGPT Apps directory listing (in progress).
Is Faxify MCP HIPAA-compliant? No. Faxify is not HIPAA compliant and does not sign Business Associate Agreements, and that applies to the MCP server exactly as it does to the apps. Do not send PHI through Faxify. Faxify's carrier and document-storage providers hold their own BAAs with Faxify, which extends no HIPAA coverage to you.
How does Faxify MCP authenticate?
Two flows. Stdio clients (Cursor, Claude Desktop) use a JWT token from your Faxify account settings, where you can also see when it expires. HTTP clients use OAuth 2.1 with Dynamic Client Registration (RFC 7591) — the client self-registers and discovers endpoints via the well-known OAuth metadata at /.well-known/oauth-authorization-server. Authorization codes are stateless JWTs encrypted with AES-256-GCM.
Can I build a custom AI fax agent on top of Faxify MCP?
Yes. Faxify MCP exposes 18 typed tools covering drafts and uploads (create, get, update, delete, list, preview, request_upload_url), sending (send, send-using-draft, resend), inbox and status (list, get, get_status, delete), and account/plans (profile, subscription, plans, offerings). The @faxify/mcp-client npm package handles the stdio-to-HTTP bridge for non-OAuth clients.
What's Next
Faxify MCP is the foundation we're building on for the next year of AI-fax work. The model that's emerging — a focused MCP server that connects a regulated, slow-moving communication channel into the fast-moving AI tools developers and operators already use — is more interesting than "fax for ChatGPT users." Once you have a typed interface to fax operations, you can build agents that do things the iOS app and the web app aren't designed for: high-volume document workflows, cross-system orchestration (fax + CRM + email in one agent), and likely patterns we haven't anticipated. We're betting that giving good tools to AI agents in a regulated channel produces workflows we couldn't have designed for in advance. We'll find out as the adoption builds.
The 5 free pages per month make it cheap to try. Setup is about ten minutes. And the 18 typed tools are enough to do real work, not just demos.
🌐 Explore Faxify MCP 📦 @faxify/mcp-client on npm · GitHub 📱 Download Faxify for iOS 📱 Get Faxify for Android 💻 Use Faxify on the Web 💰 See Faxify pricing
Written by Jangul Aslam, founder of Const Agility, LLC — creators of Faxify.