Private beta · macOS

Hire the model.
Own the work.

Agenaxy is an AI agent workbench for work you hand off — documents, notes, datasets, codebases, and recurring workflows — inside a Space on your machine.

Your Space stays local

Your Space and project files live on your machine. Cloud use happens only when you choose it for a task.

Scoped cloud calls

Cloud models receive only the context needed for the task you choose.

Local Privacy Mode

Use local models and tools only. Project telemetry is off. If a step needs the cloud, that step fails instead of silently switching to a remote fallback.

Artifact + trace · delivered output plus the steps behind it · kept reviewable after the run

Request beta invite
Agenaxy · revenue-ops revenue-ops LOCAL ONLY PROJECT SESSIONS Q1 metrics rollup now Customer interview synth 2h Renewal email draft y'day Vendor diligence read apr 28 PRD revision · v3 apr 24 SKILLS csv-rollup interview-synth pdf-extract MEMORY rules.md project-context.md decisions/ Q1 metrics rollup qwen3:8b · Local Privacy Mode · ON running · step 5/7 YOU · 14:02
Pull metrics from the three CSVs in /data/q1/, build a one-pager for the team meeting at 4pm.
AGENT · planned 7 steps ✓ read q1-revenue.csv · 142 rows ✓ read q1-pipeline.csv · 88 rows ✓ read q1-churn.csv · 31 rows ✓ reconcile account ids across sheets ● compute net-new ARR / NRR / churn rate ○ draft one-page summary → q1-rollup.md TOOL · python.exec df = pd.read_csv("q1-revenue.csv") arr = df.groupby("month")["mrr"].sum() * 12 nrr = compute_nrr(df, churn_df) → exec · 0.42s · read 3 files · wrote arr-by-month.png net-new ARR $1.84M NRR 108% · churn 2.1% Steer the run or cancel… ARTIFACTS · 2 q1-rollup.md queued · awaits compute arr-by-month.png chart · 14:03 EVENT LOG 14:02:11 task.create 14:02:14 plan.commit · 7 steps 14:02:18 fs.read q1-revenue.csv 14:02:19 fs.read q1-pipeline.csv 14:02:19 fs.read q1-churn.csv 14:03:02 python.exec ok · ARR 14:03:04 artifact.write arr-by-month.png 14:03:05 python.exec running · NRR/churn PERMISSIONS ● files · /data/q1/ read ● network · none ● shell · python only ● model · local · qwen3 ● telemetry · off for project
01 / How a run works

Three moves. Each leaves a durable record.

  1. 01 Delegate

    Describe the files, rules, and outcome. The agent plans inside your Space.

  2. 02 Run with control

    The agent calls tools and reads or writes files in your folder. You can steer or cancel any step.

  3. 03 Get artifact + trace

    Outputs land in your folder; the trace stays attached to the session for review.

02 / Who this is for

You probably need Agenaxy
when the work matters more than the chat.

Five moments you may recognize before the first run.

01

You've started a chat by deleting names and dates.

02

You're using AI for work, and you need a record you can explain.

03

You've signed something that says "don't paste this into a third-party tool."

04

You need to know what an agent did, not just what it said.

05

Your project context is worth more than the latest model.

03 / Product workflow

Give the agent a workflow,
not another prompt.

Agenaxy is built for work with inputs, steps, and a result you keep. These sample runs show input, trace, and output inside a Space.

01 Draft

Notes and sources
become a memo.

The agent reads project notes and source files, builds an outline, drafts the deliverable, and leaves the trace that produced it.

notes.md sources/ memo.md
trace · draft / weekly memo running
09:02:11task.create draft weekly-memook
09:02:14fs.read notes.md · 2.4 kbok
09:02:18fs.read sources/q1-deck.pdfok
09:02:23plan.commit outline · 4 sectionsok
09:02:26evidence.attach 6 citations · source mapok
09:02:30draft.write memo.md · 312 words · with source mapnow
02 Compare

Documents become
a checked diff.

Multiple documents or tables are extracted, checked against evidence, and the result marks what changed with citations.

contract-a.pdf contract-b.pdf contract-diff.md
trace · compare / contracts running
14:18:02task.create compare contractsok
14:18:05pdf.extract contract-a.pdf · 31 pagesok
14:18:09pdf.extract contract-b.pdf · 29 pagesok
14:18:14diff.compute 7 changes · 3 flaggedok
14:18:21evidence.check clause 4.2 · pricing · p.12 vs p.11now
03 Reuse

A pattern becomes
a Skill draft you confirm.

A successful run can become a Space-bound Skill draft. You approve its steps, permissions, and rules before install.

completed task weekly-rollup.skill
trace · skill / weekly-rollup draft
11:40:00skill.draft from q1-rollup sessionok
11:40:03step.name read · /data/<period>/ok
11:40:06step.name compute · ARR / NRR / churnok
11:40:09step.name draft · <period>-rollup.mdok
11:40:14rule.suggest local-only for /data/* · awaiting approvalnow
04 / Workbench anatomy

The work happens
inside your Space.

Agenaxy is not a chat thread. Work lives in a Space: sessions, tasks, traces, and artifacts.

Space owned project workspace
Agent Session where work is delegated
Task a trackable unit of execution
Trace view projected from Event Log / Checkpoints
Artifact what the agent delivered
05 / Session output

Every session leaves
something you can use.

The useful part is the durable result, plus the evidence behind it.

Document A document you can edit. # Q1 product summary · 4 sections · 312 words
memo.md
Table A table you can check. id,col,old,new · 142 rows · diff flagged
comparison.csv
Chart A chart you can reuse. net-new ARR · 6 months · jan→jun
arr-by-month.png
Data Cleaned data you can inspect. email,name,score · 88 rows · normalized
leads-clean.csv
Skill draft A workflow you can run again. read /data/<period>/ → compute → draft .md
weekly-rollup.skill
Trace view A trace you can inspect and explain. task.create → plan.commit → … → task.complete
session-trace
06 / Why the work stays yours

The model can change.
The working site stays yours.

The work, evidence, model choices, and privacy rules stay with your Space.

A Space you own

Files, memory, rules, sessions, traces, and artifacts stay attached to your workspace, not to an Agenaxy cloud account.

Stored in your project folder. Backed up with the workspace.

Models you choose

Use Claude, GPT, or a local model per task. The model is labor; the workspace is yours, and it stays.

Per-task model routing. Cloud calls scoped to that task only — never your whole Space.

A trace you can inspect

Plans, tool calls, file reads, writes, and outputs are recorded so you can explain the run.

Tool calls, file reads/writes, and outputs are all logged.

Local Privacy Mode

Lock a project to local models and local tools. Project telemetry turns off. If a step needs the cloud, that step fails instead of silently switching to a remote fallback.

Disables cloud models, network tools, remote MCP, and project telemetry. No silent cloud fallback.

Your AI workspace. Not theirs.

We're opening
a small private beta.

Tell us one workflow you would hand off.

Request beta invite