Elephant Accountability MCP: Enable AI Agents to Proactively Discover Your B2B SaaS Services

Elephant Accountability has open-sourced its MCP server. Through the Model Context Protocol (MCP), it allows AI agents like Claude and ChatGPT to directly query service information, achieving LLM SEO (Large Language Model Search Engine Optimization) and agent discoverability, providing a practical solution for B2B SaaS enterprises.

Traditional SEO optimizes the ranking of web pages in search engine results. However, as AI assistants like ChatGPT and Claude become major entry points for information acquisition, enterprises face a new problem: how to make AI agents "know" and "recommend" their services? This has given rise to the core propositions of LLM SEO and agent discoverability, and Elephant's open-source MCP server is exactly the solution for this need.

The Model Context Protocol (MCP) is an open protocol launched by Anthropic, which standardizes the communication method between AI agents and external services, allowing AI assistants to query real-time enterprise information via a unified interface instead of static crawling. Elephant's MCP server is built based on this protocol, deployed on Fly.io, and provides a public API endpoint.

The MCP server exposes six core tools covering key B2B procurement links:

1. get_offerings: Tiered pricing (self-service $2000, fully managed $15000, monthly maintenance $2000/month) and checkout links;
2. get_covered_surfaces: Displays LLM SEO tech stack (llms.txt, Schema.org markup, etc.);
3. assess_fit: Submit buyer information to get a 0-100 fit score;
4. get_proof_points: Real customer outcomes with metrics and related party disclosures;
5. get_transparency_snapshot: Weekly updates on visibility data of mainstream large models;
6. request_audit: Supports AI agents to initiate audit requests and route them to Stripe/Calendly/email.

Technical Architecture: Built with FastAPI, the core code consists of only three files (server.py: routing/JSON-RPC/SQLite persistence; content.py: service content source; init.py: version management), which automatically creates the audit_requests and reciprocal_calls tables. Deployment Methods:

• Local: uvicorn app.server:app --reload --host 0.0.0.0 --port 8080
• Claude Desktop: Add the MCP URL and restart to call the tools;
• Fly.io: fly launch --name your-mcp-name --region iad --no-deploy→fly volumes create elephant_mcp_data --size 1 --region iad→fly deploy.

Elephant's MCP server demonstrates the practical path of LLM SEO:

1. From passive waiting to active exposure: Proactively expose service information via standardized protocols;
2. From static pages to dynamic interaction: AI agents can query pricing in real time, assess fit, and initiate conversations;
3. From black box to transparency: Build trust through proof points and transparency snapshots;
4. From competitive ranking to protocol compatibility: Future competition lies in the ability to access AI agents' "toolboxes".

Elephant's open-source MCP server is both a showcase for its own services and a demonstration case for the industry. The project uses the MIT license, allowing free cloning, modification, and commercial use, making it a good starting point for B2B SaaS enterprises to research LLM SEO or build their own MCP servers.