Better questions. Clearer evidence.
Turn vague goals into grounded agent recommendations.
AxWise turns a goal into an evidence-bounded execution brief: who the work is for, what outcome matters, what is known or uncertain, and which capabilities the work requires. It produces operational personas, qualitative synthesis, and verified agent recommendations through a local MCP contract.
Runs locally inside Goose, Codex, Claude Desktop, or your own assistant host—with native desktop integration in Orqanix.
AxWise evaluates ambiguity, evidence sufficiency, consequence, and the value of more information. It chooses direct, evidence-assisted, bounded research, or human clarification before expensive research begins.
- Stakeholder: declared, not yet verified
- Evidence: insufficient for execution
- Candidate agents: authenticated Orqanix catalogue
The same contract works for operations, research, sales, services, compliance, or software goals. It does not assume the answer is code.
"routing_mode": "research_assisted",
"status": "pending_research",
"research_job": { "status": "queued" },
"requires_orqanix_authorization": true
}
Every recommendation is advisory. Orqanix revalidates tenant ownership, agent availability, policy, approval, budget, and tool access before execution.
From vague goal to grounded recommendation
Explore illustrative, domain-neutral scenarios. AxWise decides whether more research is valuable, resolves the customer and ideal executor, ranks existing Orqanix agents, and returns a trust-aware handoff. Orqanix's Consilium and workflow engine act downstream.
Warehouse Operations
The goal names the symptom, but the problem owner, decision-maker, evidence, and best executor still need to be resolved.
See what is quoted, inferred, or still unknown
Quote-backed persona fields retain the speaker, source document, exact quote, and character offsets. Select one of the four illustrative persona records, then hover over an extracted field to inspect its source. AxWise keeps those quotes separate from model inference and labels synthetic research as a hypothesis—not verified customer truth.
Interviewer: Lukas, how exactly do you handle raw store layout updates right now?
Lukas: Honestly, it's a completely manual mess.
We use Excel sheets to translate visual store designs in Figma into actual ordering volumes in our SAP ERP.
But maintaining 45 separate spreadsheets in Excel is incredibly labor-intensive and prone to errors.
If we make one typo, we end up over-ordering by 15% and throwing away fresh bakery goods.
AxWise advises. The host decides what may run.
Every AxWise recommendation requires authorization by the integrating host system. In the reference workflow below, Orqanix validates tenant ownership, identity, agent availability, approval, budget, and connector permissions before any external action.
AxWise intelligence, connected to Orqanix execution
The self-hostable AxWise core can enhance different agentic or workflow products; its current hosted reference API is optimized for Orqanix. The use case below shows the two products working in synergy: AxWise returns context, evidence, an execution persona, and a ranked agent or team; Orqanix plans, authorises, and runs the work.
Vague Goal
Orqanix sends the goal, tenant context, constraints, and available Agent Hub profiles.
Customer + Executor
AxWise decides whether research is needed, resolves stakeholders, and recommends the best fit.
Orqanix Validates
Orqanix checks the tenant, agent, approval policy, budget, and connector permissions.
Goal-Aware Agent
Orqanix plans and runs the work with AxWise context applied as a goal-specific overlay.
Veronika Horvat
AxWise can shape a goal-specific CFO execution persona and recommend the matching Agent Hub profile. Orqanix alone activates the agent and grants any authorised, read-only access to finance or communication sources.
Example · Orqanix-executed- 📁 Finance / Q2 Planning Drive
- 💬 Slack History (#finance)
- 📧 Email correspondence archive
Hey Veronika, I need the latest Q2 projections for the board meeting. Can you send them?
CFO AGENTHi! Pulling from Q2 Planning. Here is the file and key points:
@Veronika can you share the updated cash-flow model? I need it for the investor deck by noon.
Veronika is OOO. I found the latest version in her Finance folder:
Small prompts. Substantial work.
No special command syntax. Ask for the outcome you want in your normal chat inside Goose, Codex, or Orqanix. Your host invokes AxWise for the heavy discovery and research work.
Get a project off the ground
Make sense of what you heard
Explore another perspective
Move toward implementation
One decision layer, many kinds of work
AxWise is not limited to software delivery or product management. It can clarify the customer, stakeholder, and ideal executor for operational, commercial, research, healthcare, public-service, and technical goals.
Stakeholder Resolution
Who is this really for?
Turns a vague objective into a structured view of customers, users, buyers, affected stakeholders, desired outcomes, constraints, and unresolved assumptions.
Evidence-Aware Personas
Customer + ideal executor
Produces operational customer and executor profiles with field-level provenance, confidence, quotes, inferences, and explicit working hypotheses when evidence is incomplete.
Agent & Team Fit
Recommendation, not execution
Ranks Orqanix Agent Hub profiles against task fit, customer fit, tools, constraints, and evidence—then returns the reason, confidence, and goal-specific persona overlay.
Decisions That Improve
Outcome feedback loop
Stores immutable decision records and learns from quality, time, cost, customer reaction, and success signals without silently rewriting the evidence behind earlier recommendations.
Use the desktop. Or bring your own host.
AxWise runs as a local extension. Choose Orqanix for a complete desktop environment, or run the standalone release in Goose, Codex, or your own MCP workspace.
Included in Orqanix
The desktop bundles AxWise and its local runtime. Enable the extension when you want its specialist capabilities, and keep working in the same chat.
Orqanix adds its own sign-in, model access, and shared Results experience. Its service terms and model access are separate from the open-source license.
AxWise 0.3.0 Packages
Two launch paths, the same engine. Download the verified release packages or launch directly via terminal:
npx --yes --package=https://github.com/AxWise-GmbH/axwise-flow/releases/download/axwise-extension-v0.3.0/axwise-extension-0.3.0.tgz axwise --config /absolute/path/axwise.jsonuvx --from https://github.com/AxWise-GmbH/axwise-flow/releases/download/axwise-extension-v0.3.0/axwise_extension-0.3.0-py3-none-any.whl axwise --config /absolute/path/axwise.jsonVerified SHA-256 download checksumsExpand
a17caf788dd42e7979f281f2dce6aa3ec527ad6bd1b3a8c8a7f7b2d72c22fe20d3c13786ce62de85a8e7dc6ef380c9cb0de61ca6b6477409133b579d36fe336f- Both require Node.js 22+ and Python 3.11+ (uv can provide Python).
- Install uv before using either launch path.
- Provide your own model credentials in your local configuration; never commit secrets.
Supported and incubated by
AxWise has been supported, incubated, and mentored by leading European startup programs and deep-tech innovation hubs.
Engineers, product leaders, and architects from leading technology and enterprise organizations who starred and tested AxWise repositories:
Give every agentic goal grounded context before execution.
Self-host AxWise or embed its API in your own product and agent stack. For governed end-to-end execution, connect it to an orchestration platform such as Orqanix. Deployment control supports your security and governance programme; it does not replace it.