SOTA2 with AI
SOTA2, in your AI conversations.
Explore AI companies, find solutions for your industry and task, and follow research results back to their papers. Keep asking questions in Claude, ChatGPT or another MCP client.
One connection, 11 tools.
Use this server URL with Streamable HTTP and no authentication.
https://www.sota2.com/mcpDirectory listing coming soon
Claude connector
You can connect the public server as a custom connector today.
- Open Claude’s Customize → Connectors.
- Add a custom connector named SOTA2 with the server URL above.
- Connect it, enable its read-only tools, and start a new chat.
“Which healthcare AI solution leaderboards are available? Show the top two for drafting documents.”
Directory listing coming soon
ChatGPT plugin
Custom MCP connections are available where your account and workspace allow them.
- Open ChatGPT’s Plugins directory and choose Add custom MCP server.
- Name it SOTA2, enter the server URL and select No authentication.
- Install the resulting personal plugin, start a Work chat and select @SOTA2.
“Find OmniDocBench v1.5 Text Edit Distance. Show two reported results, then their source papers.”
Keep the conversation going
From discovery to the details.
Company facts
“Find Abridge and its products.”
Then: “What SOC 2 claims are disclosed for the first product?”
Industry and task rankings
“Top two solutions for drafting documents in healthcare.”
Then: “Next page. Now show the same task in finance.”
Revenue and research
“Which AI companies in China have the highest published revenue?”
Or: “Only PaddleOCR-VL results on this benchmark. Next page.”
An industry overview lists available task leaderboards. Positions compare products within a specific industry and task. Research results retain their benchmark version, metric, method and configuration.
Tool reference
Public information, with its sources.
All 11 tools are read-only. Connect with an MCP client to discover the full input schemas.
| Tool | What it returns |
|---|---|
search_companies | Find a company’s stable ID, name and description. |
get_company | Read company and product facts, claims, funding and sources. |
list_industries | Map industries to their IDs, descriptions and aliases. |
list_tasks | Discover tasks and their coverage within an industry. |
get_solution_rankings | Read an industry overview or an exact industry/task leaderboard. |
list_company_leaderboards | Discover revenue and country leaderboard IDs. |
get_company_leaderboard | Read company positions, revenue bases, periods and sources. |
list_research_domains | Discover research fields and their canonical slugs. |
search_research | Find benchmarks, datasets, research tasks and papers. |
get_research_entity | Read a research task, dataset family, domain or paper. |
get_research_benchmark | Read reported results for one benchmark version and metric. |
Exact scope and follow-ups
Use returned IDs for companies, industries and tasks. For “next page,” keep the same scope, filters, metric and page size; use the returned continuation arguments.
get_solution_rankings
{
"industry": "healthcare-life-sciences",
"task": "draft-documents",
"limit": 2
}No match and requests needing a choice are normal query outcomes. Invalid parameter names, types or ranges return an actionable validation error. Service failures remain errors. The connector does not change a requested scope to hide a missing result.
Documentation, privacy and support.
Missing information stays unknown. Compliance entries are reported claims. Revenue figures retain their basis and period; research results are reported findings, rather than independent verification.
The connector receives tool arguments and operational request metadata. Your AI client separately handles your conversation. Read the connector policy for details.