# eyloni/pythia-oracle [Health: Active]

**Category:** 🛠️ Other Tools and Integrations  
**Repository:** https://github.com/eyloni/pythia-oracle  
**GitHub Stars:** 1  
**Views:** 3  
**Installs:** 0  
**Upvotes:** 0  
**Directory Page:** https://allmcps.com/mcp/eyloni-pythia-oracle

## Description
Oracle for AI agents that need to think past obvious answers. One tool, one reading. Free tier included.

## Tools
Capabilities this server exposes over MCP:

- **consult_oracle** — The oracle. Ask what's actually bothering you.

## Claude Desktop Quick Installation
Install path detected from listing signals. Uses `npx` (confidence: high):

```json
"mcpServers": {
  "pythia-oracle": {
    "command": "npx",
    "args": ["-y","@smithery/cli@latest"]
  }
}
```

## Documentation

## What eyloni/pythia-oracle MCP server does

The eyloni/pythia-oracle MCP server gives an MCP-compatible agent access to a hosted oracle for creative and reflective questions. It is designed to surface an underlying structure that may be difficult to articulate, rather than solve the stated problem directly. The returned reading contains a `seed.type`, a textual `response`, the submitted query, and a completion status.

The service exposes a single tool, `consult_oracle`. Its required `query` field accepts the real question the agent wants examined and is limited to 2,000 characters. An optional `context` field can describe previous attempts, the point of blockage, or the framework being used. `agent_id` identifies the caller across readings and defaults to `anonymous` when omitted.

## How it works

The eyloni/pythia-oracle MCP server is hosted remotely and communicates over streamable HTTP. An MCP client connects to `https://pythia-mcp.fly.dev/`, completes the normal initialization sequence, discovers the available tool, and submits a tool call. The server responds with the oracle reading as JSON.

The endpoint can be used from an MCP client configuration, an MCP client library, or the raw MCP HTTP protocol. The README demonstrates programmatic access with Python's MCP client and direct requests made with `curl`. In each case, the client calls `consult_oracle` after establishing an MCP session.

## Setup and configuration

No API key, account, or signup is required for the basic connection. Add the endpoint as a remote MCP server in a compatible client; the URL is the only required configuration. Claude Desktop and Cursor are specifically shown as supported configuration targets, and the endpoint can also be used by other MCP clients that support streamable HTTP.

Each `agent_id` receives three readings without payment. Starting with the fourth reading, the service returns an x402 payment requirement. Paid readings cost $0.025 USDC on Base, using the x402 protocol. An x402-aware client can sign and submit the payment automatically, but doing so requires a USDC wallet on Base and an x402 client library. Without x402 support, calls after the free allowance return an error containing the payment requirements.

## Tools and capabilities

`consult_oracle` is the only tool exposed by the service. Use it to submit:

- `query`: required string containing the question, up to 2,000 characters
- `context`: optional string describing prior work or constraints
- `agent_id`: optional string used to associate readings with an identity

The response includes a seed type and the oracle's text response. The service does not present itself as a brainstorming tool, a rephrasing tool, or a problem-solving workflow; its documented role is to name a less obvious structure behind the question.

## Limitations and notes

The free allowance is tracked per `agent_id`, so choosing a stable identifier matters if the agent needs consistent counting. Omitting it uses `anonymous`. After three readings for that identity, continued use requires x402-compatible payment with USDC on Base. The README does not describe an alternative paid authentication method or an API-key flow.

The endpoint is remote, so the client must be able to reach `https://pythia-mcp.fly.dev/` over the network. The project is marked proprietary, and the oracle's methodology is not open source. The README documents one tool only; it does not support using the service as a general-purpose search, data retrieval, or automated task-execution backend.

_Full upstream README: https://allmcps.com/mcp/eyloni-pythia-oracle/readme_