In this category:
Table of Contents
What is MCP and why should I care?
Model Context Protocol (MCP) is an open standard that lets AI agents call tools hosted by other services. In plain terms: you point Claude (or Cursor, or any other MCP client) at our MCP server, and from that point forward you can ask questions like "pull Apex Plumbing's full carrier history" or "show me Pinnacol's 1/1/2026 Colorado filings" and the agent retrieves the data live, reasons on it, and writes it up for you.
It's the fastest way to turn your InsuranceXDate data into research, call lists, and leave-behinds without having to click through the web app for every lookup.
Before you start
- An InsuranceXDate account.
- A prepaid API balance (add funds from Settings → API). MCP calls are billed against this balance.
- An MCP-capable AI client installed locally. Claude Desktop, Claude Code, Cursor, and LM Studio are all supported.
Step 1 — Create an API key
- Go to Settings → API (or visit https://www.insurancexdate.com/settings.php#api).
- Under "Insurance Xdate API," click Create a Key.
- Give the key a name you'll recognize later (e.g. "Claude-MCP" or "Cursor-Home").
- Copy the key value and store it somewhere safe. Treat it like a password — anyone with your key can spend your prepaid balance.
The Settings page shows last-used timestamps for each key, so you can tell at a glance which keys are live and which are dormant. Rotate or delete keys you no longer need.
Step 2 — Connect your AI client
Add the following to your AI client's MCP configuration. In most clients this is a JSON file (for example claude_desktop_config.json, .cursor/mcp.json), or a "Connectors / MCP" section in the settings UI.
You can literally Tell Claude Code to do it for you from here.
I need you to connect to an MCP, the info is here - https://www.insurancexdate.com/settings.php#api - you can access with google chrome. It'll do what it needs to, asking your permission (if you set it that way) along the way. You may have to tell it to open the chrome page first, then you (human) will paste the link in there for it. It kept saying chrome wasn't active until it opened the tab, then I pasted the link and took it there.
MCP server configuration (JSON)
{
"mcpServers": {
"insurancexdate": {
"url": "https://www.insurancexdate.com/api2/McpData",
"headers": {
"X-API-KEY": "YOUR_API_KEY"
}
}
}
}
Replace YOUR_API_KEY with the key you created in Step 1. Save the file and restart the client so it picks up the new server.
Endpoint details (for troubleshooting)
URL: https://www.insurancexdate.com/api2/McpData
Protocol: JSON-RPC 2.0 over HTTP POST
Auth header: X-API-KEY: <your key>
Content-Type: application/json
Claude Code shortcut
In Claude Code you can skip editing the JSON by hand. After creating your API key, run (on one line):
claude mcp add --transport http --scope user insurancexdate \
https://www.insurancexdate.com/api2/McpData \
--header "X-API-KEY: YOUR_API_KEY"
Then quit and relaunch Claude Code (or run /exit and reopen) so it picks up the new server. You'll see the insurancexdate tools appear in the tools list once it reconnects.
Available tools and pricing
Once connected, these tools show up in your AI client. Cost is deducted from your prepaid balance as tools are called. Balance and per-tool usage are visible on the Settings page.
| Tool | Price | What it does |
| search | Free | Search the workers' comp prospect database. Returns company UIDs that feed the other tools. |
| filter | Free | Look up valid filter values: carriers, class codes, NAICS, industries, counties, agents. |
| company_details | $0.25 | Full company profile: carrier history, mod rates, premium, payroll, agents, contacts. |
| talkpoints | $0.10 | Prospecting talking points with industry and coverage research. |
| serff_search | $0.05 | Search SERFF rate filings by carrier NAIC, state, and insurance type. |
| serff_filing | $0.10 | Full filing details: narratives, coverage changes, actuarial justifications. |
Note: Data field availability varies by state. See the State Data Availability KB article for specifics on what's returned where.
The typical workflow
The tools are designed to layer. You almost always start with the two free ones and only incur cost once you know what you want.
filter — free — look up valid values you'll pass into search (carrier names, class codes, counties, etc.).
search — free — find prospects by location, class, carrier, renewal date range, and other criteria. Returns company UIDs.
company_details — $0.25 per UID — pulls the full policy, carrier history, premium/payroll, agent, contacts, and flags for a specific company.
talkpoints — $0.10 per UID — prospecting insights and coverage research tailored to that company.
serff_search — $0.05 — using a carrier's NAIC code (from company_details), find their recent rate filings in a state.
serff_filing — $0.10 — pull the full narrative and analysis on a specific filing ID.
Rule of thumb: filter/search are free, so iterate there until the result set is exactly what you want. Only then do you spend on company_details.
Use case — building a rate-shock prospect list and a leave-behind
Here's an end-to-end example of how a member used the MCP in Claude Code to turn a single account review into a targeted prospect list and a custom flyer — total cost under $2.00 in API spend.
Scenario
Agent is working an account, Apex Plumbing in Jefferson County, CO, and wants to find similar prospects and understand whether the shared carrier (Pinnacol Assurance) is pushing through a rate shock that can be pitched against.
Step-by-step
Pull the anchor account. Ask the agent: "Return all information for Apex Plumbing in Jefferson County, CO." The client runs search (free) to resolve the UID, then company_details ($0.25) for the full profile. Output: 12+ years on Pinnacol through agent Innovise, premium just jumped ~90% ($98K → $186K), xdate 6/1/2026.
Find look-alike prospects. "List 10 similar businesses (same class code) renewing in the next 60-90 days within 5 miles of Apex." The client runs another search (free) on Class 5183 / Jefferson County / renewal window. Result: only 5 in the 5-mile radius — the agent keeps it tight and asks for names on just those 5.
Enrich the shortlist. 5 x company_details at $0.25 each = $1.25. Now the agent has names, addresses, xdates, current carriers, premiums, and agents for all 5 prospects.
Spot the rate shock. "Which of these had premium increases?" No new API calls — the agent compared current and prior terms from the carrier history already pulled. One prospect, Quipco Inc., showed a +13.8% increase on flat payroll — a classic rate-only increase.
Confirm it with SERFF. "Cross-reference Quipco's carrier group and state for filings indicating rate increases." The client runs serff_search ($0.05) on Pinnacol in CO, filters to RATE_CHANGE filings flagged unfavorable to policyholders, and pulls serff_filing ($0.10) on the one that matters: filing 2528, approved 12/11/2025, effective 1/1/2026. Tier 1 policyholders getting up to +12.9%, 500+ businesses flagged for double-digit hikes, construction NAICS directly affected.
Turn it into a leave-behind. "Build a one-page infographic for Quipco." The agent gets a print-ready HTML flyer with Quipco's profile, premium history, the SERFF filing callout, and a CTA block — ready to print and drop off in the field.
What it cost
search, filter: free (used several times)
company_details on Apex + 5 prospects: 6 x $0.25 = $1.50
serff_search: $0.05
serff_filing: $0.10
Total session: $1.65
Output: a qualified 5-account prospect list with rate-shock context and a printable flyer for the hottest lead — all from one conversation.
Cost and workflow tips
Start free, then drill in. Always use search/filter to narrow before calling company_details. A sloppy search wastes $0.25 per misfire.
One key per client. If you run multiple MCP clients (say Claude Desktop at home and Claude Code at work), create a separate key for each. Makes it obvious where usage is coming from, and you can rotate or revoke one without breaking the other.
Watch the balance. The Settings page shows current month usage by tool and your remaining prepaid balance. You can set up a top-up reminder based on your typical spend.
Cache in the conversation. The AI client keeps its conversation context, so a second question about the same company doesn't need another company_details call. Anchor to one lookup, then ask multiple follow-ups in the same chat.
Use talkpoints selectively. talkpoints adds prospecting angles but it's $0.10 per call. Save it for accounts you're actually going to reach out to — not for browsing.
Common questions
Which AI clients work with this?
Any client that speaks MCP. We've tested Claude Desktop, Claude Code, Cursor, and LM Studio. If your client supports custom MCP servers with HTTP transport and a custom auth header, it'll work.
Does the MCP have access to everything my web login has?
It accesses the same underlying dataset, scoped by the permissions associated with your API key. Tools like company_details return the same data you'd see by opening a company page in the web app.
Can I use the same API key for the REST API and the MCP?
Yes. The key authenticates you against both surfaces. If you're building a script that calls the REST endpoints directly (not via MCP), use the same X-API-KEY header.
What happens if my balance hits zero?
Paid tool calls (company_details, talkpoints, serff_search, serff_filing) will start failing with an insufficient-balance error. Free tools (search, filter) continue to work. Add funds from the Settings page and calls resume immediately.
How is this different from using the REST API directly?
Same data, same pricing. The MCP is just the AI-friendly front door — it lets an LLM call the tools conversationally and chain them together. If you're writing scripts or building your own integrations, call the REST API. If you're doing research, prospecting, or ad-hoc analysis in an AI chat, use the MCP.