Part of Solutions. LeadOcean holds 693M people and 63M companies. The MCP server and the REST API draw on the same data and the same record balance.
The work you do
An agent builder ships software that acts on a goal, and the agent is only as good as the tools it can call. Three jobs repeat on every project.
- Pick the data tools. Choose what the agent can look up, with inputs and outputs it can reason about and a cost you can predict.
- Keep calls cheap and bounded. An agent loops. You need free sizing calls, hard limits on rows returned and no surprise invoice.
- Handle misses. People are missing, emails bounce, filters return nothing. The agent needs a clean "not found", not a hallucinated answer.
For the wider picture, read the API for AI agents, explained for builders. For product choices, see the best GTM agent platforms.
How teams like yours use LeadOcean
1. Register the MCP server as the agent's toolbox
The hosted server at https://api.leadocean.io/mcp has 12 tools and signs in with OAuth 2.1, so no key is pasted into a config. It supports dynamic client registration at https://api.leadocean.io/oauth/register (scope mcp), so your client can register itself.
claude mcp add --transport http leadocean https://api.leadocean.io/mcpSix of the 12 tools return data and cost records: search_leads, get_lead, find_phone, get_company, search_companies and export_leads. Six are free: count_leads, list_filters, list_enum_values, get_account, get_export and list_exports. Setup steps are on the Claude Code page.
2. Call the REST API as a function tool
If your agent runs on your own loop, wrap the endpoints as functions. POST /v1/people/search takes a flat JSON body, returns up to 100 people per page and paginates with cursor. POST /v1/people/enrich takes exactly one of linkedin_url, email, phone or person_id.
curl -X POST "https://api.leadocean.io/v1/people/enrich" \
-H "x-api-key: $LEADOCEAN_API_KEY" -H "content-type: application/json" \
-d '{"linkedin_url":"https://www.linkedin.com/in/jane-doe","reveal_email":true}'A person we hold nothing for is a 404 and costs nothing, so the agent can branch on the status code. The spec is at api.leadocean.io/openapi.json, so you can generate tool schemas from it.
3. Let the agent size before it spends
Tell the agent to count first. Send count=true as a query parameter on search with limit 1, or call count_leads. You get meta.total, capped at 100,000, and no rows. Neither costs a record.
Example agent step: "Count VPs of sales in Canada. If the total is under 6,000, pull the first page." On 30 September 2026 that count was 5,066 mailable people. Add a row limit in your tool wrapper so a loop cannot pull more than you intend.
What it costs
There are two plans. Free is 1,000 records, one-off, no card, with the same API and MCP server as Pro. Pro is $499 a month, flat. See pricing.
A record is one person or company returned. Counts, filter lookups, a 404 on enrich and an MCP get_lead that finds nobody are free. There is no per-record price, no credits and no overage invoice.
Worked example. Your agent runs 3,000 research jobs a month. Each one reads one company and returns 10 people, so 11 records. That is 33,000 records, and Pro is still $499. Rate limit is 100 requests per second per key, with 5 keys on Pro. A very large month past the fair-usage line is paced to one request a minute until the reset, not cut off.
Use the Free 1,000 to test your agent end to end. It is a pilot, not a monthly allowance.
What it does not do
LeadOcean is a data source, not an agent platform. It has no sequencer, no CRM, no dialer and no Chrome extension. It has no table or workflow builder and no agent runtime. You bring the model, the orchestration and the sending tool.
The app at app.leadocean.io has an Exports page: a filter builder, a column picker, the record price before you start, and a CSV download. Bounced emails are not refunded or credited back, so have the agent read email_status before it hands an address on.
Start here
- Connect LeadOcean to Claude Code: one command, OAuth sign-in, a free first prompt.
- Account Researcher skill: a domain in, an account brief out.
- LeadOcean + n8n: call the API from an automation node.
Comparing data sources for your agent? See the best B2B data providers for AI agents.
FAQ
Do I need an API key for the MCP server?
No. The MCP server uses OAuth sign-in, so no key is pasted. The REST API uses an x-api-key header. Both draw on the same record balance, and Free keys get search and enrich scopes.
Can my agent look up a person by name and company domain?
Not in one call. There is no name plus domain lookup. Search by domain first, then enrich by person_id. The search costs a record per person returned, so cap the page size.
How fresh is the data?
The dataset is refreshed monthly. Each record also carries its own fetched_at date, so your agent can skip stale rows.
What happens when the 1,000 free records run out?
Move to Pro at $499 a month for unlimited searches, enrichments and lookups, 5 API keys and the whole database. Your agent should handle a payment error as a stop signal, not retry.
Give your agent B2B data tools with a free API key
Free to start. No credit card. 1,000 records to spend whenever you like.
Get your free API key →