Skip to main content

Getting Started with Enginy MCP

Problem it solves

Sales reps who already use Claude or another AI tool daily have to constantly switch between their AI workspace and Enginy to run outbound. Every time they research a prospect in Claude, they have to open Enginy separately to build the list, set up the sequence, or check what's running. MCP removes that friction — Enginy works from inside the AI tool they're already in, so the full outbound motion (finding prospects, enriching, sequencing, tracking) happens without leaving the conversation. For teams building AI workflows, MCP makes Enginy a programmable part of their stack — callable from any MCP-compatible client alongside their other tools.

Use cases supported

  1. Build and enrich a prospect list from a plain-language brief. "Find 50 Sales Executives at Barcelona-based SaaS companies with 50-500 employees" → Enginy returns them enriched, ready to drop into a sequence.

  2. Run multi-step outbound plays in one workflow. Enrich, score, build campaign, write personalised messages, review before send — all from one prompt.

  3. Pull campaign performance and identify what's working. "Pull performance across all active campaigns and tell me what's driving replies" → the agent surfaces patterns across sequences, openers, segments, and send times.

  4. Triage the inbox and draft the next move. "Look at today's replies, classify each (interested / objection / out-of-office), and draft responses for my review."

  5. Chain Enginy with other tools in a single conversation. Because MCP clients can hold several connectors at once, Enginy becomes one node in a wider toolchain — e.g. pull a prospect from a meeting notes tool, enrich them in Enginy, add to a campaign, all in one conversation.

How it works

MCP (Model Context Protocol) is an open standard that lets AI tools communicate with external platforms in real time during a conversation. Once Enginy is connected via MCP, your AI tool can call Enginy directly — finding prospects, enriching contacts, building campaigns, reading the inbox — without you leaving the AI interface.
​
Learn more on our MCP product page: enginy.ai/sales-intelligence/mcp

Enginy runs a hosted remote MCP server. There's nothing to install and no API keys to manage. Authentication is OAuth 2.0, and the whole connection takes about two minutes once your workspace has MCP enabled.

Supported AI tools: Claude (web, desktop, Team, and Enterprise), Claude Code, ChatGPT (Free, Go, Plus, Pro, Business, and Enterprise), Codex, Cursor, VS Code with GitHub Copilot Agent mode, Gemini CLI, Microsoft 365 Copilot, and other clients that support remote MCP over HTTP. Claude Code is the best-supported path today — simplest setup, most testing behind it.

Enginy is available in ChatGPT as a published plugin. Personal accounts (Free, Go, Plus, Pro) and Business workspaces connect with no admin step; Enterprise workspaces need a workspace admin to enable the plugin for your role first.

Not sure whether your AI tool supports MCP? Look for a Connectors, Integrations, or MCP section in its settings — that's where these connections live in most clients.

Before you start

You'll need:

  • An active Enginy account and your login credentials

  • Your AI tool already installed

  • About two minutes

Setup has two layers, and it matters which one you're in. Step 1 is done once, by an admin. Step 2 is done by every user, and looks different depending on your tool.

If Step 1 hasn't happened, Step 2 fails — usually with a confusing authorization error, or the option to connect simply won't appear. If something's not working, check with your admin before troubleshooting your AI tool.

Step 1 — Enable MCP (once, by an Enginy admin)

  1. Go to Enginy → Integrations

  2. Find MCP server and turn on the toggle

  3. Review the connection policy:

    • Allowed redirect URIs — the callback URLs permitted to complete sign-in. Many clients use a local callback such as http://localhost:3118/callback. If a redirect URI isn't on this list, users will see redirect_uri is not allowed.

    • Allowed OAuth client IDs — leave empty during rollout so any client can connect. Once you know which tools your team actually uses, pin them.

    • Policy scopes — the maximum permissions any user is allowed to approve. Start read-only and add write access deliberately.

Tightening this policy later can invalidate connections that were already working — affected users will need to reconnect under the new limits.

Step 2 — Connect your AI tool

Learn more - https://app.enginy.ai/mcp

Claude — Team or Enterprise plans

On Team and Enterprise plans, only an organization Owner can add a custom connector — individual members can't add it themselves.

Owner, once:

  1. Go to Organization settings → Connectors

  2. Click Add, choose Custom

  3. Enter the name Enginy and the server URL: https://openapi.enginy.ai/mcp

  4. Click Add

Each member, after that:

  1. Go to Customize → Connectors

  2. Find Enginy (labeled "Custom") and click Connect

  3. Sign in with your Enginy credentials and approve the requested scopes

Everyone else — Claude (Free, Pro, Max), Claude Code, ChatGPT (Free, Go, Plus, Pro, Business, and Enterprise), Codex, Cursor, VS Code, Gemini CLI

These all follow the same shape: no extra admin step beyond Step 1 — you connect and authorize yourself. The Enginy server URL is the same everywhere: https://openapi.enginy.ai/mcp

Claude (Free, Pro, Max)

  1. Go to Settings → Connectors, or straight to claude.ai/settings/connectors

  2. Click "+" next to Connectors and select "Add custom connector"

  3. Paste the server URL and save

  4. Authenticate with your Enginy credentials when prompted

Claude Code

This is a terminal (command-line) tool, run by whoever already has Claude Code installed — not something typed into a chat window.

Run:

claude mcp add --transport http --callback-port 3118 enginy <https://openapi.enginy.ai/mcp>

Then type /mcp in Claude Code and complete the browser approval flow.

Microsoft 365 Copilot - Business

Available on a qualifying Microsoft 365 subscription — this includes both the free Copilot Chat tier and the full paid Microsoft 365 Copilot add-on. A Microsoft 365 admin sets this up once for the company; you then authenticate individually the first time you use it.
​
​Admin, once:

  1. Go to the Microsoft 365 admin center → Copilot → Connectors

  2. Add a new connector, name it Enginy, and provide the server URL: https://openapi.enginy.ai/mcp/reduced

Each user, after that: the first time you try to use Enginy inside Copilot Chat, Word, Excel, or another Copilot surface, you'll be prompted to sign in with your Enginy credentials and approve the requested permissions.

ChatGPT — Free, Go, Plus, Pro, Business, and Enterprise

Personal accounts and Business workspaces: no admin step is required.

  1. In ChatGPT, go to Settings → Plugins (or open the Plugins tab in the sidebar) and search for “Enginy.”

  2. Click Install plugin, then Sign in with Enginy.

  3. Complete the Enginy sign-in and permission screens.

  4. Enginy will now appear as an available plugin in your chats.

Enterprise workspaces: your workspace admin must first enable the plugin and its underlying app for your role. If you don’t see Enginy listed, contact your workspace admin.

The plugin automatically includes Enginy’s Skills Library—no separate installation is required.
​

Codex

Same idea — a terminal command, for whoever already has Codex installed.

Add the server:

codex mcp add enginy --url <https://openapi.enginy.ai/mcp>

Start the sign-in flow:

codex mcp login enginy

Complete the browser approval when prompted.

Cursor

Go to Settings → Tools & MCPs and click "New MCP server" to open mcp.json.

Add:

{"mcpServers":{"enginy":{"url": "<https://openapi.enginy.ai/mcp>"}}}

Project-level config lives in .cursor/mcp.json; global config in ~/.cursor/mcp.json.

Restart Cursor or refresh MCP servers, then approve Enginy in the browser.

VS Code / GitHub Copilot Agent mode

Add to .vscode/mcp.json:

{"servers": {"enginy":{"type": "http","url": "<https://openapi.enginy.ai/mcp>"}}}

Gemini CLI

Also a terminal tool. Open ~/.gemini/settings.json and add:

{"mcpServers":{"enginy":{"url":"<https://openapi.enginy.ai/mcp>","oauth":{"enabled":true}}}}

Restart Gemini CLI, then authenticate:

/mcp auth enginy

A browser window opens for Enginy sign-in. Once you approve, run /mcp to confirm Enginy is connected. This needs a local browser — it won't complete over SSH or on a headless machine.

Other remote MCP clients

Add https://openapi.enginy.ai/mcp in the client's MCP settings and follow its remote MCP instructions. A client can connect to Enginy if it supports remote Streamable HTTP MCP and OAuth 2.0 with PKCE. If your client only speaks stdio, you'll need a bridge such as mcp-remote in front of it.

Confirm it worked

Ask your AI tool to run mcp_whoami. It returns your Enginy user, your workspace, and the scopes you've been granted — confirming both that the connection is live and what it's allowed to do.

If it returns nothing, or only a handful of tools appear where you expected many, the usual cause is a narrow approved scope set or a stale cached authorization, not a broken connection. See Troubleshooting.

What you can do

Around 150 operations across the platform: contacts, companies, lists, campaigns, AI Research / AI Message / AI Snippets, AI Finder, inbox, tasks, and analytics.

Some actions still require the Enginy UI: setting team roles and permissions, connecting Identities, and sender warmup.

For more advanced use cases, explore Enginy's Skills Library (skills.enginy.ai) — pre-built prompt bundles for specific outbound workflows. If you're connecting via the ChatGPT plugin, the entire Skills Library is already included.

Known limitations

  • Each user authenticates individually, under their own Enginy permissions. On Claude Team and Enterprise, an org Owner adds the connector once, but every member still connects and authorizes for themselves — there's no way to push a live connection to a team in one action.

  • Not available in the Gemini app — Gemini CLI is a separate developer tool, not a substitute for the same daily use.

  • Complex multi-step workflows may need more than one prompt. Specific requests route better than vague ones.

  • MCP operations consume the same credits as the Enginy UI. There's no separate MCP charge. Paid operations appear in the Activity tab.

  • Tool calls are logged with user attribution and timestamps.

Troubleshooting

My AI tool keeps asking me to reconnect, or says a permission is missing.

This looks like a permissions problem but almost never is. The client is re-running the full OAuth flow on each attempt instead of reusing the token it already has — and each reconnect registers a new client, which keeps the loop going.

Don't keep clicking Connect. Instead:

  1. Fully remove the Enginy connector from your AI tool — remove, not just disconnect

  2. Re-add it once and complete a single authorization

  3. Run mcp_whoami to confirm the session is live before trying anything else

  4. Still looping? Try a different client or surface (Claude desktop vs. web) to isolate whether it's specific to one build

A real permission rejection returns a specific message — Tool requires scopes: ... or Requested scopes are not allowed — not a silent reconnect loop.

I get redirect_uri is not allowed or OAuth client_id is not allowed.

These come from workspace policy, not your setup. Ask your Enginy admin to add the exact redirect URI (including port and path) or client ID to the allowlist in Integrations → MCP.

No tools appear, or only a few.

Your approved scope set is narrower than the tools you're trying to use, or the client cached stale auth state. Run mcp_whoami to see your granted scopes, then reconnect to re-approve with a wider set if your workspace policy allows it.

If an action seems to fail, or a tool seems to be missing, after connecting via Microsoft 365 Copilot — check that the connector's server URL ends in /mcp/reduced, not just /mcp. This is the most common cause.

The agent used the wrong tool, or couldn't find one.

Be specific in your prompt. The tool surface is broad and auto-generated, so vague requests route poorly. "Use AI Finder to search for VPs of Finance at Series B SaaS companies" beats "find me some prospects."

Something else

Full technical reference and error table: docs.enginy.ai/mcp/security-troubleshooting

Security & permissions

On first connection, you'll be asked to approve the scopes MCP is allowed to perform. After that, MCP is scoped to your own Enginy permissions — if you can't see something in the Enginy UI, the agent can't access it via MCP either.

Tool results are returned to the AI tool you've connected. This means your data passes through your AI provider (Claude, Codex, Gemini, etc.) in the same way it would if you were chatting with them directly. Data retention on the AI provider side is governed by your own agreement with that provider.

To revoke access at any time: go to Enginy Integrations → MCP and disable the connection, or remove Enginy from your AI tool's connector settings. Tokens are invalidated immediately.

Did this answer your question?