AxWise MCP
A local Rust MCP server for product discovery: research plans, synthetic interviews, interview analysis, PRDs and delivery briefs.
- 62 ★ on the repository
- Last commit
- Apache-2.0
Extended access
Why this level
- The server installs as a binary, writes into the project workspace and optionally uses your model API keys.
Install
- Works in
Manual install
codex plugin marketplace add AxWise-GmbH/axwise-flow --ref axwise-codex-v0.1.1Then `codex plugin add axwise@axwise`; for Apple Silicon macOS.
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- The server installs as a binary, writes into the project workspace and optionally uses your model API keys.
This is third-party code. Review the repository files before installing.
What it does
AxWise ships as a single Rust executable `bin/axwise` and runs as a local MCP server for product discovery. It builds a research plan, selects market evidence, creates synthetic personas and interviews, analyzes interviews and writes a PRD and a delivery brief. Python, Node and a separate subscription are not needed. By default (`--model-access host`) credentials and model choice stay with your agent. If the client supports MCP sampling it receives bounded text-generation requests, otherwise the candidate goes through `advance_artifact` and the current model writes and reviews the draft itself. Optionally an OpenAI, Anthropic or Gemini key can be passed through the environment. There is a plugin for Codex on Apple Silicon.
Who it is for. Product managers and founders who hand an agent the path from hypotheses to a PRD and want to see which parts are assumptions and which are data.
Good fit when
- A pilot needs a research plan and a proposed PRD.
- User interviews must be analyzed and requirements derived.
- Synthetic personas are needed to test hypotheses in advance.
- A delivery brief is assembled from an already approved PRD.
Not a fit when
- Real interviews and surveys are needed: synthetic ones do not replace them.
- The environment is not Apple Silicon macOS: the prebuilt binary targets only that.
- The agent does not follow model requests and bypasses `advance_artifact`.
Example request
Use AxWise to create a draft PRD for a cat-food pilot in Estonia with three shops. Label assumptions and do not invent research.MCP
- Transport
- stdio
- Authentication
- not required
Environment variables
AXWISE_MODEL- A model granted by your provider for API key mode
OPENAI_API_KEYsecret- Optional OpenAI key; Anthropic and Gemini keys are read the same way
Limitations
The prebuilt binary targets Apple Silicon macOS, is ad-hoc signed and not notarized, while other systems use a source build through cargo. The tool timeout in the client must be at least 180 seconds. Synthetic personas and interviews serve as hypotheses and do not replace research with people, and an agent must not present such data as real. The host must carry an artifact through to a confirmed receipt and not substitute its own text for server output. Verification is described for Codex CLI with a regular ChatGPT login.
How to disable. Remove the server from the client, for example with `codex mcp remove axwise`, and the folder of the unpacked release. Clean the project workspace by hand.
FAQ
How do I install the Codex plugin?
Run `codex plugin marketplace add AxWise-GmbH/axwise-flow --ref axwise-codex-v0.1.1`, then `codex plugin add axwise@axwise`. The plugin targets Apple Silicon macOS.
How do I connect to another MCP client?
Configure the same absolute path and arguments as a stdio server with a tool timeout of at least 180 seconds.
Who pays for the model?
Your agent and its subscription by default. Key mode is enabled with the `--model-access api-key` flag.
What do synthetic interviews do?
They create hypotheses to test and do not replace research with real people.
How do I build from source?
Run `cargo build --release --locked --manifest-path crates/axwise-core/Cargo.toml --bin axwise`.
Discussion
No questions or reviews yet.
Related
MCP serversA Rust manager that brings MCP services under one daemon: a CLI, a web UI, an HTTP API and SDKs for Python and Rust
MCP server of the self-hosted Synaplan AI platform: semantic search over your documents, long-term memory and chat
A single Rust gateway between an agent and all its MCP servers: the client sees 9 to 17 meta-tools and finds the rest on demand.
A local web service and MCP server that gathers several models to answer, peer review each other or debate as advisor personas.
A Rust host for coding agents: a dashboard, required checks, review by another model and a signed receipt for each turn
A plugin that gives an agent the IntelliNode skill and MCP server: ask any model, search, review code, tests, SQL, images and speech
README in short
The README covers the server purpose, Codex plugin installation, installation in Codex CLI, model access through the agent or a key, environment variables and modes.