ProspectingConnector
Clay
Push rows into Clay tables, run Clay routines and search its people and company data from Claude or ChatGPT via Tulina's MCP server, on your own Clay credits.
Official site: www.clay.com
What it can do
3 read/1 write
- 01Write
Write up to 50 rows per call into a Clay table through its webhook
- 02Read
Run a Clay routine, a function or workflow, on up to 100 items
- 03Read
Search Clay's people and company database
- 04Read
Read rows from existing tables on Clay's Enterprise plan
Clay is where many teams enrich a list. The Tulina connector lets an agent feed that list: write rows into a Clay table, run a Clay routine on a batch, or search Clay's database, then carry the result on to the CRM, the sequencer and the inbox in the same conversation. Every call spends Clay credits on the account concerned, exactly as the same work would in Clay.
Does Clay have an API?
Yes, in two forms. Clay's developer documentation describes a public API with searches, routines, tables (Enterprise only), audiences and signals. Clay tables can also receive rows through an inbound webhook, which is how most teams push data in today. Clay also documents HTTP API columns, which call other tools from inside a table; that is the opposite direction. Tulina's connector covers the webhook, routines, search and table reads.
Does Clay have an official MCP server?
Yes. Clay documents a hosted MCP server that connects its workspace to Claude, ChatGPT, Microsoft Copilot and Glean. According to Clay, admins control which reps have access, a per-rep credit limit, and which functions are exposed, and work done through MCP draws on the same Actions and Data Credits as work in Clay, with no separate MCP charge. Clay's MCP page also mentions finding contacts, getting emails and phone numbers, and returning up to 1,000 people search results. It is a strong route when Clay is the only system the model needs.
How do you connect Clay to Claude and ChatGPT?
- For writing rows, open the table in Clay, choose Add, then Monitor webhook, and copy the cURL command it shows.
- For routines and search, create a Clay API key in Settings, then Account, then API keys.
- In Tulina, open the Clay card. Add each table by pasting its cURL command (the URL and token are picked up), and the API key as a second entry. Each table entry has a name that the agent uses to target it.
- Add
https://mcp.tulina.ai/mcpto Claude or ChatGPT and sign in with your Tulina account. The MCP page has the same link.
Clay has no API to create a table webhook, so each table is added from Clay's interface, once.
Tulina's Clay connector and Clay's own MCP server
| Clay's MCP server | Tulina's Clay connector | |
|---|---|---|
| Who runs it | Clay, hosted | Tulina, with your Clay entries held in the vault |
| Sign-in | A login to the Clay workspace, per rep | Table webhooks and a personal Clay API key, held by the person or the organisation |
| Writes to a table | Not stated on the pages read | clay_push_rows, up to 50 rows per call, with a receipt |
| Runs a routine | Functions exposed by an admin | clay_run_routine on 1 to 100 items, then clay_get_run |
| Searches the database | People and company search | clay_search, in filters or query mode |
| Next to your other tools | Clay | Clay plus the CRM, enrichment, sequencing and inbox connectors |
If Clay is the only system in play, Clay's own server is the shorter path. Tulina earns its place when the list Clay builds is one step in a longer piece of work. Compare the economics in the Clay pricing breakdown and Clay vs Apollo.
What you can ask it
"Send these 40 companies to the enrichment table in Clay." The agent previews the batch as a dry run, then writes the rows through the table's webhook and returns a receipt of how many succeeded and which failed. Clay then runs the table's enrichments on each new row.
"Run our company-research routine on these 20 domains." The agent starts the routine, which is asynchronous, polls the run until it completes and returns the outputs. Clay has no endpoint that lists routines, so you give the routine's id.
"Find 25 heads of sales at European SaaS companies in Clay and put them in a table." The agent runs a Clay search, pages through the results, and writes them to a table or the CRM. Searches spend Clay credits.
What runs on its own: Enrich contacts with email and phone and Outbound from buying signals show the steps around an enrichment table. Other enrichment connectors: Apollo and FullEnrich. More on the category in AI prospecting tools.
What the Clay connector can't do
- A table webhook accepts 50,000 sends in total, even if rows are deleted. Tulina counts what it sends and warns as the limit approaches; past it, you create a new webhook in Clay and paste it back on the same entry.
- Reading existing tables works on Clay's Enterprise plan only.
- One row is one JSON object. A JSON array makes a single row.
- Clay does not expose an endpoint that lists routines, so the agent needs the id. Audiences and signals, which Clay's developer documentation describes, are not in this connector.
- A timeout while reading Clay's answer may still have written the row; sending it again can create a duplicate in the table.
- The connector makes no price promise: each call spends Clay credits, and a rate limit per workspace returns a wait time.
Why through Tulina
The connection was never the hard part.
Most systems now publish their own way for a model to reach them, and that is good: it is what Tulina connects to. The hard part is deciding who may use it, what it is allowed to return, and what the model knows before it asks.
The keys stay in one place
YouTeamOrganisation
Connect Clay once. The credential is encrypted in a vault, never held in clear, and never pasted into a chat or a workflow file. A person uses their own key where they have one, their team's where they do not, the organisation's after that, so nobody is blocked and nobody holds more access than they need. Revoke it once and everything running on it stops.
What leaves is filtered before it leaves
fields droppedmasked or hashedpseudonymised
Roles decide who can reach Clay at all. Beyond that, a policy on the connector itself decides what comes back. The data is minimised at the source rather than after it has already reached a model.
Every call is on the record
whatwhowhen
Runs are logged for the organisation. That is what makes an automated action reviewable after the fact, and it is what a security review asks for first.
The run starts knowing your company
namingkey accountsdecisions
A direct connection gives a model access. It does not tell it how your company names things, which accounts matter, or what was decided last quarter. Tulina carries that across every tool, so the same request against Clay returns something usable instead of something generic.
Processes that run on Clay
None published yet.