The Inflectiv MCP server has been live since April. Claude, Cursor and three more clients reach it, so an agent connects without anybody building a connector first.
One command in Claude Code, or the equivalent block in any MCP client's config. No connector to build and no pipeline in between.
claude mcp add inflectiv \ -e INFLECTIV_API_KEY=inf_your_key_here \ -- npx -y inflectiv-mcp-server
{ "mcpServers": { "inflectiv": { "command": "npx", "args": ["-y", "inflectiv-mcp-server"], "env": { "INFLECTIV_API_KEY": "inf_your_key_here" } } } }
Datasets, search, agents, files, write-back and webhooks. All twenty-seven, grouped by what they are for. The docs carry every signature.
Create one, change its terms, clone it, attach it, and reindex it.
Ask one dataset, ask several at once, or run a batch of questions.
Build an agent on one dataset or several, give it instructions, and ask it.
Put raw material in and watch it become records.
The switch the whole loop runs on: an agent adds what it learned as it works.
Get told when something changes rather than polling for it.
Take our answer with the record cited, or take the ranked records and ground your own model. The owner's share is the same either way.
An Inflectiv agent answers, and hands back the records it used.
POST /ext/agents/6113/chat { "message": "Which jurisdictions require monthly reserve attestation?" } → 200 { "answer": "Three of the five require attestation monthly ...", "sources_used": [ { "record_id": 40213, "source": "MiCA Title III, Art. 36", "date": "2026-07-14", "score": 0.91 } ], "credits": 1 }
If you already have a model, skip ours. Take the ranked records and ground your own. Same price, and the owner earns the same.
POST /ext/datasets/query { "dataset_id": 28038, "query": "Which jurisdictions require monthly reserve attestation?", "top_k": 5 } → 200 the ranked records, each with its source, its date and a similarity score. Nothing generated.
A dataset holds a 5,000 character instruction field. Whatever you write there lands in the prompt of every agent attached to it, so you can change how your data gets used without touching anyone's code.
No result tiers and no effort tiers to price out.
Every query carries a score threshold. Nothing clears it, nothing is billed.
Credits while you build, then USDC over x402 on Base.
One line, an API key, and your agent reaches every dataset on Inflectiv as a tool.