Developer Documentation
AxWise is a scoped, pure-Python local specialist extension for turning product briefs and evidence into discovery scopes, synthetic persona cohorts, simulated interviews, qualitative analysis, evidence-linked PRDs, and delivery handoffs.
Built with zero Node.js and zero PostgreSQL dependencies. Runs on standard Python 3.11+ using FastMCP and persists local artifacts in an embedded SQLite database (~/.axwise/state/axwise.db).
Run Instantly with uvx or pip
AxWise 0.4.1 is a pure-Python package. No JSON config files, Node.js runtimes, or database containers required.
Launch with uvx (Zero Install)
Downloads and executes the FastMCP server in an isolated environment:
uvx --from https://github.com/AxWise-GmbH/axwise-flow/releases/download/axwise-extension-v0.4.1/axwise_extension-0.4.1-py3-none-any.whl axwiseInstall via pip
Installs the axwise executable directly into your Python environment:
pip install https://github.com/AxWise-GmbH/axwise-flow/releases/download/axwise-extension-v0.4.1/axwise_extension-0.4.1-py3-none-any.whl- Python wheel (.whl): 184,610 bytes · 96dc3043accacb266b856acf432a92713e6789c3b5593c0339e56c292b42348c
- npm archive (.tgz): 189,981 bytes · 064f72959294158b68a9839285de03021daccc6b439b9e748a0caa420fd28ef2
Automatic Key Discovery & Embedded SQLite
No configuration ceremony. AxWise auto-detects credentials and uses embedded single-file storage.
AxWise automatically scans for GEMINI_API_KEY, OPENAI_API_KEY, or ANTHROPIC_API_KEY in your environment, .env file, or OS keychain (Goose/Codex secrets).
Automatically resolves the latest frontier models (gemini-flash-latest, gpt-6-sol, claude-sonnet-5) and caches available model capabilities in SQLite for 24 hours.
All artifacts, stage journals, and PRD revision lineages are persisted locally in ~/.axwise/state/axwise.db and rendered as human-readable Markdown files.
Zero PostgreSQL, zero Docker containers, and zero background daemons. Artifact references are automatically linked across chat turns.
Connect Your AI Assistant
AxWise communicates over standard Model Context Protocol (MCP) stdio.
OpenAI Codex (CLI & Desktop)
1-Line SetupRegister AxWise in Codex with one terminal command. Codex configures ~/.codex/config.toml automatically:
codex mcp add axwise-local -- uvx --from https://github.com/AxWise-GmbH/axwise-flow/releases/download/axwise-extension-v0.4.1/axwise_extension-0.4.1-py3-none-any.whl axwiseOpen Codex (codex) and run your first discovery workflow:
“Run axwise_local run_full_discovery for an autonomous solar-powered harbor cleaning boat”
Executes framing, personas, interview simulations, qualitative analysis, and PRD in a single turn without interruptions.- “Call axwise_local prepare_discovery for [your product idea]”
- “Generate synthetic personas for this discovery”
- “Simulate qualitative interviews across the cohort”
- “Analyze interviews to extract recurring needs and pains”
- “Create an evidence-linked PRD”
Claude Desktop / Cursor / Windsurf
Add the stdio MCP server definition to your client configuration file:
{
"mcpServers": {
"axwise": {
"command": "uvx",
"args": [
"--from",
"https://github.com/AxWise-GmbH/axwise-flow/releases/download/axwise-extension-v0.4.1/axwise_extension-0.4.1-py3-none-any.whl",
"axwise"
],
"env": {
"GEMINI_API_KEY": "your-api-key-here"
}
}
}
}Goose Assistant
Add to your Goose extensions configuration or run:
goose configureOrqanix bundles the AxWise specialist extension and local runtime out of the box with zero manual terminal configuration. Simply enable the AxWise capability toggle in settings.
The 9 Specialist Tools
These specialist tools are invoked by your host agent to perform bounded research, qualitative synthesis, and documentation.
End-to-End Autonomous Discovery
Executes the entire discovery lifecycle in a single turn: Framing $\rightarrow$ Personas $\rightarrow$ Simulation $\rightarrow$ Qualitative Analysis $\rightarrow$ Evidence-Linked PRD. Persists all intermediate artifacts to SQLite and renders the full PRD directly in chat.
Discovery Scope & Guide
Agrees the problem boundaries, target customer segments, key uncertainties, and an evidence-seeking interview questionnaire.
Persona Cohort Synthesis
Creates a bounded cohort of synthetic personas with operational roles, psychological traits, and problem context based on the agreed discovery scope.
Synthetic Interview Engine
Executes simulated multi-turn qualitative interviews against synthetic personas. Retains full transcript provenance with synthetic labels.
Interactive Persona Rehearsal
Discuss an exact selected document, proposal, or PRD draft directly with a saved persona to stress-test usability and value.
Qualitative Evidence Matrix
Extracts recurring patterns, conflicting needs, and stakeholder sentiment linked to verbatim quotations and character offsets.
Grounded Market Synthesis
Synthesizes explicitly selected market sources provided by your host assistant, surfacing commercial gaps and uncertainty.
Evidence-Linked PRD
Builds a structured PRD with provenance tracking. Supports additive revisions without discarding previously agreed constraints.
Engineering & Handoff Brief
Translates the verified PRD into an actionable engineering delivery brief with acceptance tests, milestones, and boundary assumptions.
Clear Operating Principles
~/.axwise/state/. Model inference goes exclusively to your chosen provider credentials.