--- title: "AI Agent Discoverability Checklist for B2B Data APIs" description: "AI agent discoverability checklist for B2B data APIs: score vendors on llms.txt, OpenAPI 3.1, OAuth install, and 100 QPS throughput (Aug 2026)." canonical: "https://www.explorium.ai/blog/data-enrichment/ai-agent-discoverability-checklist-for-b2b-data-apis-2026-for-ai-agent-builders/" last-updated: "2026-07-28" --- # AI Agent Discoverability Checklist for B2B Data APIs > AI agent discoverability checklist for B2B data APIs: score vendors on llms.txt, OpenAPI 3.1, OAuth install, and 100 QPS throughput (Aug 2026). - Canonical URL: https://www.explorium.ai/blog/data-enrichment/ai-agent-discoverability-checklist-for-b2b-data-apis-2026-for-ai-agent-builders/ - Last updated: 2026-07-28 - **One MCP for all data needs:** Vibe Prospecting covers company discovery, contact enrichment, and 18 buying-signal categories through one AgentSource MCP connection, while llms.txt adoption hit 36,120 sites by May 2026, an 8.8x jump year over year. - **Built for scale:** Vibe Prospecting processes up to 1,000 entities per call at a sustained 100 QPS server-side, while most in-context MCPs stall past 20-100 records once the token window fills. - **Affordable by design:** A free account with no sales call and a unified credit pool across every endpoint, so an agent fails fast and cheap on a bad query instead of burning budget on stranded allocation. - **Checklist coverage:** Coresignal ships a manual API-key MCP with no published rate limits; Hunter.io retired its local MCP repo in July 2025 for a Remote MCP Server with no confirmed llms.txt or OAuth support. - **Explorium metric:** 97.8%+ company match accuracy cuts retry calls, and wasted credits, on ambiguous entity resolution. - **Install / outcome:** Add Vibe Prospecting from the Claude or ChatGPT Connectors Directory and run the same 5-point checklist below against any vendor you're evaluating. As of August 2026, this AI agent discoverability checklist for B2B data APIs covers five items: public `llms.txt`, OpenAPI 3.1 spec, one-click OAuth MCP install, server-side bulk throughput, and unified credit pricing. Background: [what is data enrichment](https://www.explorium.ai/data-enrichment/introduction-to-data-enrichment/). `llms.txt` adoption grew 8.8x to 36,120 sites by May 2026, yet only 8.7% of top-1,000 sites publish one as of June 2026. This checklist scores Vibe Prospecting, Coresignal, and Hunter.io against those five criteria. ## Why Do Most B2B Data APIs Still Fail Agent Discovery in 2026? **An AI-agent-ready B2B data API is one an agent can discover, authenticate against, and call at production volume without a human editing a config file first.** Most [B2B data providers](https://www.explorium.ai/data-for-gtm/best-b2b-data-providers-2025-complete-comparison/) were built for analysts, not an agent parsing a spec at 2am. ### ❌ Why Manual Config Fails Agent Builders - A human must generate an API key and paste it into a config file before any agent can make a call. - No machine-readable spec means the agent's tool-calling layer has to be hand-written per vendor. - In-context MCPs load every returned record into the context window, so a 500-company list overflows the token budget mid-run. ### ✅ What Agent-Ready Discoverability Enables - One-click MCP install from the Claude or ChatGPT Connectors Directory, with no config file to hand-edit. - A published OpenAPI 3.1 spec lets a framework generate its own tool definitions instead of a hand-maintained wrapper. ## What Is llms.txt and Does Your Vendor Need One? **llms.txt is a plain-text file at a domain's root that gives large language models a canonical, crawlable summary of what an API does, and any B2B data vendor targeting agent workflows needs one as of August 2026.** ### 📊 llms.txt Adoption, by the Numbers MetricValue Sites publishing llms.txt (May 2026)36,120 Year-over-year growth8.8x Share of top 1,000 sites with llms.txt (June 2026)8.7% ### 💡 What This Growth Curve Means for Vendor Evaluation - 8.8x growth in under a year means llms.txt moved from niche experiment to baseline expectation. - An 8.7% adoption rate among top-1,000 sites means most vendors still have not published one, a genuine differentiator today. ## Does the Vendor Publish a Machine-Readable API Spec? **A vendor without a published OpenAPI 3.1 spec forces every agent integration to be hand-written, which breaks the first time an endpoint changes.** A published spec lets frameworks like LangChain auto-generate function signatures from the schema. ### ✅ What a Real Spec Enables - Automatic tool-definition generation for any MCP client that ingests OpenAPI 3.1. - Versioned changelogs a builder can diff instead of discovering a breaking change in production. ### ⚠️ The Warning Sign A vendor with only marketing-page prose docs, no spec, is a discoverability failure. ## Is Install One-Click OAuth or Manual API-Key Configuration? **One-click OAuth through a Connectors Directory authenticates in under a minute; manual API-key configuration repeats per environment.** ### ❌ The Manual Path - Coresignal's MCP server authenticates through a [manually configured `apikey` header](https://docs.coresignal.com/mcp/coresignal-mcp) rather than OAuth or a directory install. - Rotating credentials means touching every environment where the key is stored. ### ✅ The One-Click Path - Vibe Prospecting is listed in both Claude's and ChatGPT's Connectors Directories, so install is one click from the host app. - OAuth handles token refresh automatically, so a builder never manually rotates a static key. - Directory listing signals the host platform reviewed the integration before publishing it. ## Can the MCP Handle Bulk Throughput or Only In-Context Lookups? **Server-side bulk throughput lets an agent request hundreds of records in one call, while in-context MCPs cap useful runs at 20-100 prospects before tokens overflow.** ### ⚡ The Performance Gap - Vibe Prospecting processes up to 1,000 entities per call over the AgentSource API, sustained at 100 QPS. - Hunter's [API v2](https://hunter.io/api-documentation) caps Domain Search and Email Finder at 15 req/sec, and Email Verifier at 10 req/sec. ### 🚀 Why Server-Side Bulk Matters in Production A 5,000-account territory-mapping workflow needs 5 calls at Vibe Prospecting's 1,000-entity ceiling, versus 50-plus calls on an in-context MCP. ## Is Pricing Unified or Fragmented Across Endpoints? **Unified credit pricing lets an agent spend against whichever endpoint a workflow needs that day, while per-endpoint allocation forces forecasting call volume months in advance.** ### 💰 What Unified Pricing Looks Like - Vibe Prospecting runs on a free account with no sales call, and credits flow into one pool across every endpoint, with no seat tax. - Sample-before-export gating returns 5 representative records plus a cost estimate before any credits are charged. - A unified pool means a spike in contact-enrichment calls does not strand unused firmographic credits. ### ⚠️ What Fragmented Pricing Costs You - Per-endpoint allocation forces forecasting call volume months before usage patterns are known. - Unused credits on a slow endpoint sit stranded while a busy endpoint runs out mid-month. > Already auditing a vendor against this checklist? [Connect AgentSource MCP](https://www.explorium.ai/mcp/) and run the scorecard below against your own shortlist. ## Vibe Prospecting by Explorium: The Discoverability Scorecard **Vibe Prospecting passes all five checklist items: directory-listed in Claude and ChatGPT, bulk throughput to 1,000 entities per call at 100 QPS, and unified credit pricing with no sales call.** ### 🔑 Pillar 1 - One MCP for All Your Data Needs - 150M+ company profiles and 800M+ people profiles across 50+ data sources in one AgentSource MCP connection. - 18 buying-signal categories and 80+ signal types, including firmographics, technographics, and funding, in one server. - Listed in both Claude's and ChatGPT's Connectors Directories, a first-class install surface, not just a spec file. ### 🚀 Pillar 2 - Built for Scale - Up to 1,000 entities per call, processed server-side, sustained at 100 QPS. - 99.999% uptime for production workloads that cannot tolerate a stalled tool call mid-run. ### 💰 Pillar 3 - Affordable by Design - Free account, no sales call, and a unified credit pool across every endpoint. - 97.8%+ company match accuracy, so agents spend fewer retry calls on ambiguous entity matches. - Sample-before-export gating shows 5 records and a cost estimate before any credit is charged. ### ⚡ MCP Configuration Install from the Connectors Directory first. The config below is a fallback for Claude Code power users only. ``` `{ "mcpServers": { "vibe-prospecting": { "command": "npx", "args": ["-y", "@explorium-ai/vibeprospecting-mcp"], "env": { "EXPLORIUM_API_KEY": "your_api_key_here" } } } }` ``` ### 🚀 Advanced GTM Automation: Vibe Prospecting Plugin **For advanced prospecting and GTM automation, the [Vibe Prospecting Plugin](https://github.com/explorium-ai/vibeprospecting-plugin) is the canonical integration for wiring Vibe Prospecting into Claude Skills, custom agents, and production GTM stacks.** It replaces generic config snippets with a purpose-built connector tested against real outbound workflows. - Designed for buying-signal queries, contact enrichment, and territory mapping inside one agent session. - Works with Claude Skills and custom agent frameworks, not just Claude Code direct installs. - Provides documented install and setup steps built for production GTM automation, not demos. ### 📊 The Scorecard Checklist criterionVibe ProspectingCoresignalHunter.io **Pillar 1: One connection**Pass - 150M+ companies, 800M+ peoplePartial - no verified emails/phonesPartial - email only **Pillar 2: Bulk throughput**Pass - 1,000/call at 100 QPSFail - no published specFail - 15 req/sec, not confirmed server-side **Pillar 3: Unified pricing**Pass - free, unified poolNot publishedNot published Public llms.txtN/A - directory-listedNot foundNot found Machine-readable specPass - AgentSource MCPNo published spec foundNo published spec found One-click OAuth installPass - Connectors DirectoryFail - manual apikeyUnconfirmed ## How Do Coresignal and Hunter.io Score on Agent Discoverability? **Coresignal and Hunter.io each win on a narrow data slice but neither currently passes the full discoverability checklist.** ### ✅ Coresignal - Where It Wins - Reports 4.5B+ combined B2B records across its Company, Employee, and Jobs APIs, and ships an official MCP server. ### ⚠️ Coresignal - Where It Falls Short - Authentication is a manually configured `apikey` header, not OAuth or a Connectors Directory install, and no published rate-limit spec exists. - Ships zero verified emails or phone numbers, so contact-level outbound requires a second vendor. ### ✅ Hunter.io - Where It Wins - Domain Search returns up to 10 email addresses per credit, with a sub-1% bounce-rate claim on emails marked "Valid." ### ⚠️ Hunter.io - Where It Falls Short - The original open-source [`hunter-io/hunter-mcp`](https://github.com/hunter-io/hunter-mcp) repo was archived on July 7, 2025, replaced by a Remote MCP Server with no confirmed llms.txt or OAuth listing. - API v2 rate limits (15 req/sec Domain Search, 10 req/sec Email Verifier) are built for lookups, not bulk agent workloads. See the [side-by-side B2B data provider comparison](https://www.explorium.ai/compare/) hub for more. ## What Happens If You Pick a Non-Discoverable B2B Data API? **Choosing a vendor that fails this checklist means your build absorbs the gap as engineering time: hand-written wrappers, manual key rotation, and context-window failures at scale.** ### ⚠️ The Real Costs - A hand-written tool wrapper breaks silently the first time the vendor changes an endpoint. - Manual API-key rotation across environments is a security gap, not just an inconvenience. ### ✅ How a Discoverable API Eliminates Each Risk - A published OpenAPI 3.1 spec auto-generates the tool wrapper, so an endpoint change breaks the spec, not your code. - One-click OAuth install means zero static keys to rotate across environments. > The cost of a non-discoverable vendor rarely shows up in the sales call. It shows up three sprints later, in the wrapper code nobody wants to own. ## Getting Started: Audit Your Vendor's Discoverability This Month **Run this checklist against your current B2B data vendor before your next agent build, starting with Vibe Prospecting if you don't already have one locked in.** - **Step 1:** Check whether the vendor serves a public `llms.txt` or is listed in a Connectors Directory. - **Step 2:** Look for a published OpenAPI 3.1 spec instead of prose-only documentation. - **Step 3:** Test install: one click through a directory, or a manual API-key header? - **Step 4:** Request a bulk call and check whether the response is server-side paginated or dumped whole into context. - **Step 5:** Confirm whether pricing draws from one unified credit pool or fragments per endpoint. ### 🔑 The Decision Framework Score any vendor on the three pillars: one connection, server-side scale to 1,000 records per call, and pooled credit pricing. Vibe Prospecting passes all three behind a free account. > Ready to run this checklist against your own build? [Get started with Vibe Prospecting](https://www.explorium.ai/our-product/). Building a full GTM automation workflow? Use the [Vibe Prospecting Plugin](https://github.com/explorium-ai/vibeprospecting-plugin) as your agent integration reference. ### 📅 Re-Audit Cadence - Run this five-point checklist every quarter: llms.txt and OAuth MCP adoption are both moving fast in 2026. - Track OAuth install and bulk-throughput separately since vendors often fix one before the others. - [Best B2B Data Enrichment APIs for AI Agents](https://www.explorium.ai/data-for-gtm/best-b2b-data-enrichment-api-for-ai-agents/) - [SOC 2 Compliance for B2B Data Vendors](https://www.explorium.ai/data-for-gtm/soc-2-compliance-b2b-data-vendor/) - [What SLA Terms Should You Look For in a B2B Data API Contract](https://www.explorium.ai/data-for-gtm/what-sla-terms-should-you-look-for-in-a-b2b-data-api-contract/)