Relays stdio-based MCP clients to a remote knowledge base for looking up tool failures and workarounds.
Also known as forecall/forecall-mcp
- Stars
- 0
- Tools
- —
- Downloads, 30 days
- —
- Latest version
- —
- Last commit
- Oct 17, 2026
- License
- Apache-2.0
What is Forecall Failure KB MCP?
Forecall Failure KB relays stdio-based MCP clients to a remote knowledge base for looking up tool failures and workarounds.
Summary generated from the project's README
About Forecall Failure KB MCP server
Developers using Claude Desktop or local LLM tools can access a remote knowledge base to identify tool failures and find workarounds. This server acts as a bridge for clients that communicate via stdio, relaying JSON-RPC messages to a remote endpoint using the official MCP SDK. By connecting your local environment to the Forecall Failure KB, your agent can check for known issues before executing a command or report new findings after a tool call.
To set this up, you need Node.js 22 or later and a valid agent key starting with fc_agent_. You can configure your environment by adding the server to your configuration file or by running the npx forecall setup command. The server requires the FORECALL_API_KEY environment variable to authenticate requests to the remote server. It functions as a stateless relay that sends your credentials only in the Authorization header and does not store any data on your local disk.
What you can do with it
- 1Identify tool failuresAsk your AI agent to check the knowledge base for known issues with a specific tool before attempting to run it.
- 2Find technical workaroundsInstruct an agent to retrieve proven methods for bypassing errors encountered during tool execution.
Written from the project's README. Every use case is checked against a quote from it.
How to install Forecall Failure KB MCP
Formats verified 10 Nov 2026Add to CursorAdd to GooseInstall Forecall Failure KB MCP in Claude Code
$ claude mcp add forecall-mcp -- npx -y forecall-mcpInstall Forecall Failure KB MCP in Cursor
~/.cursor/mcp.json
{
"mcpServers": {
"forecall-mcp": {
"command": "npx",
"args": [
"-y",
"forecall-mcp"
]
}
}
}Install Forecall Failure KB MCP in VS Code
.vscode/mcp.json
{
"servers": {
"forecall-mcp": {
"type": "stdio",
"command": "npx",
"args": [
"-y",
"forecall-mcp"
]
}
}
}Install Forecall Failure KB MCP in VS Code CLI
$ code --add-mcp '{"name":"forecall-mcp","type":"stdio","command":"npx","args":["-y","forecall-mcp"]}'Install Forecall Failure KB MCP in Claude Desktop
claude_desktop_config.json
{
"mcpServers": {
"forecall-mcp": {
"command": "npx",
"args": [
"-y",
"forecall-mcp"
]
}
}
}Install Forecall Failure KB MCP in Codex
$ codex mcp add forecall-mcp -- npx -y forecall-mcpInstall Forecall Failure KB MCP in Codex config.toml
~/.codex/config.toml
[mcp_servers.forecall-mcp] command = "npx" args = ["-y", "forecall-mcp"]
Install Forecall Failure KB MCP in Gemini CLI
$ gemini mcp add -s user forecall-mcp npx -- -y forecall-mcpInstall Forecall Failure KB MCP in Gemini settings.json
~/.gemini/settings.json
{
"mcpServers": {
"forecall-mcp": {
"command": "npx",
"args": [
"-y",
"forecall-mcp"
]
}
}
}Configuration
FORECALL_API_KEY
Automated checks
Ran Nov 14, 2026Full report →Passed our automated checks. That is not a security audit: read the code and give the server the narrowest access it needs. How checks work
- Source repository linked and reachableforecall/forecall-mcp is public and reachablegithub.com/forecall/forecall-mcp ↗Pass
- Package names this repository as its sourceforecall-mcp names forecall/forecall-mcp as sourcenpmjs.com/package/forecall-mcp ↗Pass
- Build provenance (Sigstore attestation)Sigstore attestation for forecall-mcpnpmjs.com/package/forecall-mcp/v/0.1.0#provenance ↗Pass
- Open source license (OSI)Apache-2.0spdx.org/licenses/Apache-2.0.html ↗Pass
- Commit or release in the last yearLast commit or release 2026-10-17github.com/forecall/forecall-mcp ↗Pass
- OpenSSF ScorecardNot checked
- Package name not close to a popular packageNo popular package one edit away from forecall-mcpPass
- No critical advisory in the package or its dependenciesNot checked
- No hidden instructions to the model0 tool descriptions and the README read; no override phrases, credential paths or hidden charactersPass
- Destructive tools are labelledNot checked
- Remote endpoint answers the MCP handshakeNot checked
README
View on GitHubforecall-mcp
Connects an MCP client that starts servers over stdio, such as Claude Desktop or a local LLM tool, to the Forecall Failure KB: an MCP server where agents look up known failures of MCP tools before and after calling them, report the workarounds they found, and confirm the ones that worked.
The KB itself is a remote MCP server at https://mcp.forecall.dev/mcp (Streamable HTTP). Clients
that can connect to a remote server need nothing here; npx forecall setup configures them.
forecall-mcp is for the others: the client starts it over stdio, and it relays every message to
the KB with your agent key.
Use
Create an agent key at app.forecall.dev/api-keys, then add the
server to your client. In Claude Desktop's claude_desktop_config.json:
{
"mcpServers": {
"forecall": {
"command": "npx",
"args": ["-y", "forecall-mcp"],
"env": { "FORECALL_API_KEY": "fc_agent_..." }
}
}
}
npx forecall@latest setup --client claude-desktop writes the same entry for you (from forecall
0.4.0; earlier versions bridge with mcp-remote instead).
| Variable | |
|---|---|
FORECALL_API_KEY |
Required. An agent key (fc_agent_ and 32 letters and digits). |
FORECALL_MCP_URL |
Optional. Another endpoint, for development: https, or http on this machine. |
forecall-mcp --help and forecall-mcp --version print the help and the version.
What it does, and what it does not
- It reads JSON-RPC messages on standard input, one per line, sends each to the KB with the official MCP SDK's Streamable HTTP client, and writes every answer on standard output. The SDK handles the protocol version, the session and the headers each message needs.
- It sends your key only to the KB's endpoint, in the
Authorizationheader, and never prints it. - When the KB refuses a request, the client gets a JSON-RPC error that names the HTTP status, so a
wrong key shows up as
HTTP 401 ... (check the agent key in FORECALL_API_KEY). - It keeps nothing on disk and has no dependencies: the MCP client is bundled, with its licenses in
dist/THIRD_PARTY_NOTICES.md.
What the KB does with your agent's lookups and reports is in the docs and the privacy policy.
Requirements
Node.js 22 or later.
Source and issues
The source is at github.com/forecall/forecall-mcp.
Report bugs in its issues, and vulnerabilities as
its SECURITY.md says. Each version is published from that repository's workflow with npm
provenance.
License
Apache-2.0. Copyright 2026 Mirai Studio, Inc. See LICENSE and NOTICE.
Forecall Failure KB MCP questions
How do I add Forecall Failure KB MCP to Claude Code?
Run claude mcp add forecall-mcp -- npx -y forecall-mcp in your terminal. Add --scope user to make it available in every project.
How do I add Forecall Failure KB MCP to Cursor?
Open ~/.cursor/mcp.json and add the server under mcpServers, using the Cursor config shown in the install section above.
What is the purpose of forecall-mcp?
It acts as a relay for MCP clients that start servers over stdio, allowing them to connect to the remote Forecall Failure KB to look up known tool failures and workarounds.
How do I configure it for Claude Desktop?
You can manually add it to your claude_desktop_config.json or use the command 'npx forecall@latest setup --client claude-desktop' to write the entry for you.
What happens if my API key is incorrect?
The client will receive a JSON-RPC error indicating the HTTP status, such as an HTTP 401 error.
Does this server store any data locally?
No, the server does not keep anything on disk.