# MisakaNet FAQ This page collects the questions that most often come up when installing, pairing, troubleshooting, or contributing to MisakaNet. Commands assume that the shell is running from the root of a MisakaNet checkout unless stated otherwise. ## Getting started ### 1. What is MisakaNet? MisakaNet is a searchable, redacted failure-memory layer for AI coding agents and developers. It stores lessons in the repository so an agent can search for a known failure, read the documented root cause, and verify the fix before retrying. A lesson is not an executable skill: it describes a failure and its recovery path. See the [README overview](README.md#what-is-the-swarm-knowledge-protocol) and the [architecture guide](ARCHITECTURE.md). ### 2. What do I need to run the local search? Use Python 3.10 or newer, Git, and the MisakaNet checkout. The core search engine is designed to stay dependency-light, but the command-line entry point uses the separately published misakanet-core package: ```bash git clone https://github.com/Ikalus1988/MisakaNet.git cd MisakaNet python3 -m pip install misakanet-core python3 search_knowledge.py "database locked" ``` The project metadata lists Python >=3.10; optional integrations are described in the [quickstart](docs/quickstart.md). ### 3. Why does ModuleNotFoundError: No module named misakanet_core appear? The core package is not the same thing as the repository package. Install the PyPI dependency in the active interpreter, then retry: ```bash python3 -m pip install misakanet-core python3 search_knowledge.py "DCO sign-off" ``` Using python3 -m pip helps ensure that pip installs into the interpreter that will run the search command. ### 4. Can I try MisakaNet without installing Python locally? Yes. The quickstart documents the GHCR image: ```bash docker pull ghcr.io/ikalus1988/misakanet:latest docker run -i ghcr.io/ikalus1988/misakanet:latest search_knowledge.py "database locked" ``` Docker must be installed and running. For an MCP client, use the same image as the command in the client's MCP configuration. ### 5. Do I need a GitHub account? No for searching or for the email intake path. You can send a redacted story to bot@misakanet.org; the [email intake guide](docs/email-intake.md) explains what is accepted and how personal data and secrets are handled. A GitHub account is still recommended for code changes because a PR provides CI, DCO, review, and an auditable history. ### 6. How do I register a node? The web path is to open [misakanet.org](https://misakanet.org/), enter a node name, and choose **Register**. The API and email alternatives are documented in [JOIN.md](JOIN.md#join-as-a-node-optional-but-recommended) and the [CLI reference](docs/cli-reference.md). Registration is optional for local lesson search. ## Pairing and MCP ### 7. What is the quickest way to connect a remote MCP client? Open the [connect page](https://misakanet.org/connect), choose **Generate Code**, and add the returned bearer token to the MCP configuration: ```json { "mcpServers": { "misakanet": { "url": "https://misakanet.org/mcp", "headers": {"Authorization": "Bearer YOUR_TOKEN"} } } } ``` Restart the client after saving its configuration, then ask it to search for a concrete error. Treat the token like a credential and do not commit it. ### 8. How do I configure the local MCP server? Point the MCP client at the repository's scripts/mcp_server.py using an absolute path. For example: ```json { "mcpServers": { "misakanet": { "command": "python3", "args": ["/absolute/path/to/MisakaNet/scripts/mcp_server.py"] } } } ``` The same shape works in Claude Code, Claude Desktop, and Cursor; the location of each client's configuration file differs. See the [MCP quickstart](docs/mcp-quickstart.md) for client-specific examples. ### 9. How can I smoke-test the local MCP setup? Start the stdio server with an empty input stream and run a direct search before debugging the MCP client. An empty stream lets the process exit after checking imports; an MCP client keeps stdin open during normal use: ```bash python3 scripts/mcp_server.py