mcpField
O
OpenAny

OpenAny

Remote
openany.siAI & Agents1Updated 1d ago

About OpenAny

OpenAny is a remote MCP server for a marketplace offering access to AI models and callable APIs through one API key and prepaid balance. Agents can inspect the API catalog and call APIs; requests may incur per-call charges.

Key features

  1. 01Access models and APIs with one key
  2. 02Discover available MCP tools
  3. 03Search and inspect APIs on the lite endpoint
  4. 04Execute APIs with per-call billing

Use cases

  1. 01Find an API for an agent task
  2. 02Review an API schema and price before use
  3. 03Call a selected API with spending approval

MCP support

Provides a remote MCP server over Streamable HTTP. Tool discovery is public; tool calls require an OpenAny API key, and API execution is billed per call.

Configuration

Remote servers need no local install. Pick your client, then copy the config or install in one click.

Endpointhttps://mcp.openany.si/mcp
Copy install command
Claude CodeTerminal
Cursor.cursor/mcp.json
VS Code.vscode/mcp.json
Claude Desktopclaude_desktop_config.json
Windsurf~/.codeium/windsurf/mcp_config.json

Run commands in a terminal; paste JSON into the client’s MCP config and restart. Servers that need sign-in will prompt on first connect.

All endpoints
Streamable HTTPhttps://mcp.openany.si/mcpAuth required

Tools

Details

From the project README

OpenAny MCP

OpenAny (https://openany.si) is the marketplace for agents: one key for 168+ AI models and 3,200+ APIs. No subscriptions, no per-provider signups. Top up once, pay per call. This repo is OpenAny's remote MCP server (Streamable HTTP).

Endpoint: https://mcp.openany.si/mcp

Transport: Streamable HTTP, not legacy SSE.

Operator: Alephant AI LLC.

Pricing: Pay per call, no subscription. Check the current resource price and billing unit before approving a paid call.

OpenAny provides model access and callable APIs under one API key and prepaid balance. The supplied catalog snapshot is approximately 168+ models, 3,201 APIs and 3,220 MCP tools. These are approximate, changing catalog figures supplied for this documentation, not independently verified live counts or a guarantee that every model is exposed as an MCP tool. Use the current tools/list response and tool schemas to determine MCP availability. The separate model API base URL is https://api.openany.si/v1.

Optional lightweight endpoint

When enabled and routed by the deployment, https://mcp.openany.si/mcp-lite uses the same Streamable HTTP transport and exposes just three tools: search_apis, get_api and run_api. It lets an agent search the API catalog, inspect one API's schema and pricing, then execute the selected API without loading the full independent-tool catalog into context.

The implementation enables this interface by default; operators can disable it with MCP_LITE_ENABLED=false. Public deployment availability is a separate check. search_apis and get_api read metadata; run_api forwards a business request and may incur charges, requiring the caller's gateway credentials and spending authorization. Configure one endpoint, not both, unless your workflow explicitly needs both catalogs. See the lightweight guide. The examples and registry descriptor in this repository continue to use /mcp.

Authentication

  • tools/list is public; discovery does not require an OpenAny API key.
  • Resource execution through tools/call requires an OpenAny API key. The gateway remains responsible for authentication, billing and execution.
  • No OAuth: this server does not implement an OAuth sign-in flow.
  • Confirmed authentication header: Authorization: Bearer <OPENANY_API_KEY>. Supply an OpenAny API key through OPENANY_API_KEY; the examples use this confirmed header format.

The header format is confirmed by the maintainer for this documentation update. This is separate from validating a particular account, client version or business request. No new authenticated live call was performed during this update; see the verification record for test scope.

Provide the key through OPENANY_API_KEY in the client's private environment or secret settings. Never hardcode a key in the repository, paste it into an agent prompt, or include it in a URL, screenshot, log or support message. Ensure the client process can read that environment variable; a desktop app does not necessarily inherit a terminal's environment.

Client setup

Merge the appropriate snippet into existing settings rather than replacing unrelated servers. The files below contain environment references, not real keys. Every example uses the confirmed Bearer header format; none is a claim of an end-to-end paid test of every client.

ClientExampleDestination
Claude Codeclaude-code.shRun from a private shell with OPENANY_API_KEY available
Cursorcursor.mcp.json~/.cursor/mcp.json
Codexcodex.config.toml~/.codex/config.toml
Pipi.mcp.json~/.pi/agent/mcp.json or .pi/mcp.json
Hermeshermes.config.yaml~/.hermes/config.yaml
OpenClawopenclaw.json~/.openclaw/openclaw.json, under mcp.servers
Other clientsgeneric.mdClient-specific remote MCP settings

JSON does not support comments. The JSON examples use a top-level $comment annotation for the warning and CLI notes; only the mcpServers or mcp block is connection configuration. If your client rejects annotation fields, copy that configuration block without $comment, retaining the warning separately.

The requested Claude Code command expands the environment variable before invoking its CLI. The resulting header can be visible to local process inspection and may be stored in personal client configuration. Do not run it with shell tracing or commit the generated configuration. The Pi/OpenClaw CLI alternatives in annotations are alternatives, not additional commands to run as a batch. OpenClaw's CLI alternative sets the endpoint and transport only; the environment-backed header still needs to be merged from the JSON example.

Client reference documentation: Claude Code, Cursor, Codex, Pi, Hermes, OpenClaw. Client versions and configuration support can change; consult the matching client documentation if a field or command is rejected.

Verify without accidentally spending

  1. Connect to the full endpoint URL using Streamable HTTP.
  2. List tools and inspect the intended tool's current schema. Discovery alone does not prove authentication, balance access or successful execution.
  3. Before a live call, review the current price, parameters and total spending limit, and obtain explicit approval. Configure the key privately.
  4. Confirm the result of the authorized tools/call, its gateway status and available usage or billing record. Do not equate HTTP 200 with business success. Never repeat an uncertain paid request automatically.

No new authenticated live tools/call was performed during this documentation update.

Registry metadata

server.json is the descriptor for the official MCP registry's 2025-12-11 schema. Its required Authorization header describes the confirmed authenticated-call configuration, not a credential requirement for public tools/list.

The registry description is limited to 100 characters, so its wording is shorter than the full description above. Version 0.1.0 is published on the official MCP Registry as si.openany/openany. Local setup does not run mcp-publisher publish.

Development

The existing Go server implementation, configuration, migrations and tests are preserved. See development and operator notes, the lightweight connection guide and existing verification records. The optional lightweight interface is separate from the full /mcp endpoint documented here.

Links

License

MIT. Copyright (c) 2026 Alephant AI LLC.

Comments

Markdown supported

New to MCP?

From what the protocol is to installing your first server, these answers cover most first questions.

An open protocol that defines how AI applications connect to external tools and data in a uniform way. Clients talk to the model; MCP servers provide tools, resources and prompt templates. They exchange standard messages, so one server can be reused by many clients.