Key takeaways:
- Apollo.io’s MCP server runs at
https://mcp.apollo.io/mcpover Streamable HTTP with OAuth 2.0, and needs no API key and no local install. - It is unusually complete for a GTM server, covering prospect search, enrichment, contacts, lists, sequences, one-off email sends, tasks and analytics.
- Apollo prohibits AI model training on data reached through MCP, and asks you to turn training off in your AI client before connecting.
- People search returns profiles, but emails and phone numbers require a separate enrichment action that consumes credits.
A quick disambiguation first, because two well-known products share this name. This review covers Apollo.io, the B2B prospecting and sales engagement platform, and its MCP server at mcp.apollo.io.
This review is not about the Apollo MCP Server from Apollo GraphQL, which connects AI agents to GraphQL APIs and is a completely different product for a completely different audience. Search results for “apollo mcp” mix the two freely, and Palantir ships an unrelated Apollo as well. If you arrived here wanting GraphQL tooling, this is the wrong page.
With that settled: the Apollo.io MCP server is one of the more capable GTM servers available, and it is worth understanding what it does before you decide whether it is the right one for your stack. We publish a competing sales engagement platform with its own MCP server, so we will be explicit about where we differ and where Apollo is genuinely ahead of us.
What Apollo MCP Is
The Apollo.io MCP server exposes the platform’s prospecting database and engagement features as tools an AI client can call. It is a remote, hosted server, so nothing runs on your machine and there is no package to install.
The connection details are refreshingly plain:
| Setting | Value |
| Endpoint | https://mcp.apollo.io/mcp |
| Transport | Streamable HTTP |
| Authentication | OAuth 2.0 |
| API key required | No |
| Local install required | No |
Apollo offers three routes in. Built-in connectors cover ChatGPT, Claude, Perplexity, Replit and Cursor, where you connect an account without touching configuration. First-party coding plug-ins cover Codex and Coworker. The standalone server is for anything else that supports remote MCP over Streamable HTTP.
The three-route structure is a good signal of maturity. A vendor maintaining directory listings, coding plug-ins and a documented standalone endpoint has committed engineering to the protocol rather than shipping a wrapper over an existing API. Partners building on Apollo register an OAuth app before connecting, which is a separate flow again.
Being clear about what an MCP server is not matters here, because the two surfaces get conflated. MCP is the interactive layer for agent-driven work, while a REST API remains the right tool for bulk imports, scheduled jobs and webhooks. That division holds for us too, which is why we maintain a public API alongside our server rather than treating one as a replacement for the other.
That OAuth-only stance is a deliberate design choice rather than a missing feature, and it follows where the specification has gone. The MCP authorization specification requires authorization servers to implement OAuth 2.1 and requires MCP servers to implement OAuth 2.0 Protected Resource Metadata for discovery. Apollo implementing OAuth rather than a bearer key is the specification-aligned path.
What Apollo MCP Can Do
This is where Apollo’s server separates itself. Most vendor MCP servers cover one job. Apollo’s covers the full prospecting-to-outreach arc, which is unusual and is the strongest argument for using it.
| Capability | What you can ask for | Notes |
| Search and prospect | People and companies by title, headcount, location, industry, technology | Returns profile and availability data |
| Enrich people and companies | Verified emails, phone numbers, firmographics | Consumes credits |
| Manage contacts | Create and update contacts, check for existing records | Deduplication depends on you asking first |
| Manage lists | Create contact and account lists, add records | Groups records for sequences and views |
| Create and improve sequences | Draft multi-step sequences, rewrite for a persona, generate alternate first touches | Can be created inactive for review |
| Manage sequence enrollment | List mailboxes, add or remove contacts from sequences | List email accounts first to pick a sender |
| Send one-off emails | Individual sends with custom To, Cc and Bcc | Recipients can be non-contact addresses |
| Research companies | Company details, headcount, technologies, job postings | Hiring signals included |
| Analyze performance | Reply rates, engagement trends, sequence comparisons | Draws on Apollo analytics |
The sequence-authoring tools are more sophisticated than the category norm. You can ask Apollo to build a five-step outbound sequence for a named persona, keep step one under sixty words, and leave the sequence inactive so a human reviews it before anything sends. You can also ask it to generate three alternate first-touch emails using different angles, such as a pain point, a trigger event and social proof.
There is a two-step pattern worth internalizing, because it catches people out. People search returns profiles and availability, not contact details. Retrieving an email address or phone number is a separate enrichment action, and that is the step that draws down credits. Budgeting for a prospecting workflow means counting enrichments, not searches.
That split is standard across the category rather than an Apollo quirk. Most AI prospecting APIs price discovery cheaply and charge at the point of resolution, because finding that a person exists is far less expensive than confirming how to reach them. Reading the pattern correctly is what stops a conversational workflow burning a quarter’s credits in an afternoon.
How to Set Up Apollo MCP
Before connecting, confirm four things: an active Apollo account, access to the features and records you want reachable, available credits for enrichment, and model training switched off in your AI client.
That last item is a requirement rather than a suggestion. Apollo prohibits AI model training with Apollo MCP integrations and asks you to verify the setting before connecting, which is a meaningful contractual position given that the data is licensed B2B contact information.
Claude Code
Add the server with a single command, then authenticate:
claude mcp add --transport http apollo https://mcp.apollo.io/mcpRun /mcp in Claude Code and complete the OAuth flow in your browser.
If you manage several Apollo workspaces, such as separate client accounts, register the same endpoint under different names and authenticate each one separately:
claude mcp add apollo-work --transport http https://mcp.apollo.io/mcp
claude mcp add apollo-personal --transport http https://mcp.apollo.io/mcpBoth connections then coexist in one session, and you direct a request to a specific account by name.
OpenAI Codex
Add Apollo to ~/.codex/config.toml:
[mcp_servers.apollo]
url = "https://mcp.apollo.io/mcp"Authenticate with codex mcp login apollo and complete the OAuth flow.
VS Code With GitHub Copilot
Create .vscode/mcp.json in your workspace:
{
"servers": {
"apollo": {
"type": "http",
"url": "https://mcp.apollo.io/mcp"
}
}
}Open the Command Palette, run MCP: List Servers, start the Apollo server, and complete OAuth when prompted.
Claude Desktop and Other Clients
In Claude Desktop, open settings, go to Connectors, and add https://mcp.apollo.io/mcp. For any other MCP-compatible client, add the same URL as a remote server, choose Streamable HTTP as the transport, and authenticate with OAuth 2.0. Some clients label that transport as HTTP or Streaming HTTP, which is the same thing.
Workflow tools count as clients here as well. Automation platforms that speak MCP can call Apollo as one step among many, which suits scheduled prospecting better than a chat window does, and we have written up the same pattern using n8n’s MCP support. The trade is that you lose the conversational refinement that makes the sequence-authoring tools worth using.
Apollo MCP Pricing
The MCP server has no separate price. It draws on the Apollo plan and credit balance you already have, so the cost question is which tier you need and how much enrichment you run through it.
| Plan | Price per seat, per month, billed annually | Credits per seat, per year |
| Free | $0 | 900, granted monthly |
| Basic | $49 | 30,000, granted upfront |
| Professional | $79 | 48,000, granted upfront |
| Organization | $119, minimum 3 seats | 72,000, granted upfront |
Annual billing saves 24% against monthly, and Basic and Professional both offer a 14-day trial. Credits are the number to watch rather than the seat price, because enrichment through MCP consumes them exactly as it does in the app.
We should concede something here rather than talk around it. Apollo has a genuine free tier that includes API access, and we do not: we run a 14-day free trial and no permanently free plan. For a solo founder testing whether an MCP prospecting workflow is worth building at all, that difference is decisive and it favors Apollo.
The seat model is the thing to model carefully at team scale, though. Credits are granted per seat, so a five-person team on Professional buys five allocations whether or not all five prospect, and Organization enforces a three-seat minimum. Teams that find the arithmetic unfavorable at their headcount tend to end up comparing Apollo alternatives on total cost rather than on list price, which is the more useful comparison.
