One workspace · Every kind of AI work

Chat, code,automate.One living workspace.

Think, ship code, and run browser work in one place. Pick any model, run locally or in the cloud, and let each project's context update as the work changes.

Join the waitlist See how it works

Free and open source · Desktop beta September 2026

catalog / code working

SESSION / 04Add product search

GPT Codex · Local

You

Add search without changing how filters work.

Henaworking across your project

Readuse-product-filters.ts✓

Editproduct-search.tsx+42

Testcatalog.test.ts42/42

Context updateCatalog filters use URL state

Verified + saved

Continue this session...↑

01 Three ways to work

Every kind of AI work.
One workspace.

Choose Chat for thinking, Code for shipping, or Claw for recurring web work. Every project keeps its sessions, tools, and evolving context together.

Interactive product previewSwitch projects or models ↓

Hena

Product strategy · Personal project

Choose your modelSwitch anytime. Keep the project.

Chat context updates as you workHena updates notes, decisions, and research as the conversation develops.

Run on

Launch brief from interviews

You

Turn these interview notes into a launch brief.

Hena

I am reading the interviews for the problem people repeat, rather than treating every feature request as its own requirement. That keeps the brief anchored in lived friction.

Research agent2 tasks completed

Search project notes6 interviews · 18 excerpts

Done

Trace repeated problemsContext handoffs · setup work

Done

The recurring friction is rebuilding context between tools and conversations. People are not asking for more places to work; they are asking not to lose the thread.

The strongest signal is continuity: people want to pick up yesterday's work without another setup pass. I am testing that signal against team size and workflow so the audience stays specific.

Signals agent2 tasks completed

Cluster interview themesContext · decisions · follow-up

Done

Compare team patternsSmall teams · handoffs · momentum

Done

That gives the launch a clear audience: small teams moving from research into delivery. Continuity connects the insight to a daily behavior.

I will lead with one project holding its notes, decisions, and next questions together. The message should explain the change in working rhythm before it lists capabilities.

Brief agent2 tasks completed

Draft launch briefAudience · problem · message

Done

Check brief against signalsPromise · proof · next question

Done

The brief is ready: keep project knowledge together, give each project the right tools, and keep the user in control. It guides the launch without pretending the research answered everything.

Launch brief

Audience: Small teams working across research and delivery.

Message: Start with what the project already knows.

Add organization context

main

You

Add organization context to every agent run.

Hena

I am tracing a run from session creation through instruction assembly before changing the lookup. The important boundary is where organization context joins project context, not the final prompt string.

Code explorer2 tasks completed

Read session lifecyclesession setup → instruction priority

Done

Inspect context lookuporganization → project → repository

Done

The project and organization are both available before instructions are composed. The change can stay local to context assembly and preserve the existing instruction priority.

The organization guidance belongs first, with project guidance next and repository rules as the final override. I am keeping the lookup explicit so a missing organization remains a valid path.

Implementation agent2 tasks completed

Update context loader+42 / -8

Done

Add isolation tests+44 / -4

Done

Missing organizations still run normally, and switching scope cannot reuse another organization's context. The implementation is intentionally small: +86 / -12 across the change.

The boundary cases are covered. I am running the meaningful regression cases now, then I will summarize what the diff changes and what it leaves alone.

Test runner2 tasks completed

bun test37 passed

Done

Check regression pathsclean

Done

37 tests passed and the regression paths are clean. The organization context change is ready for review: +86 / -12, with the scope boundary now explicit.

Changes verified

Files changed: 4
Diff: +86 −12
Tests: 37 passed

Selected excerpts · 4 files

src/agents/context-loader.ts+30 -6
src/agents/run-context.ts+12 -2
tests/context-isolation.test.ts+30 -2
tests/context-loader.test.ts+14 -2

src/agents/context-loader.ts@@ context assembly @@- const project = await loadProject(scope)+ const organization = await loadOrganization(scope)+ const project = await loadProject(scope) const repository = await loadRepository(project)+ return [organization, project, repository]

tests/context-isolation.test.ts+30 -2

it("keeps organization context isolated", async () => { const context = await loadContext({ organizationId: "missing" }) expect(context.organization).toBeUndefined()})

tests/context-loader.test.ts+14 -2

it("does not reuse another organization", async () => { expect(await loadContext(scope)).not.toEqual(otherContext)})

Verified37 passed

02 Purpose-built project modes

Different work.
The same workspace.

01 / Chat project

Think in conversation.

Keep research, decisions, and conversations together without code-specific tooling.

02 / Code project

Work inside the codebase.

Give every session the repository, files, terminal, and review tools it needs to ship.

03 / Claw project

Hand off repetitive web work.

Keep browser tools, running tasks, and reviewable results inside an autonomous project.

03 Living project context

Context that grows
with the work.

Hena does more than carry a transcript forward. It follows changes in the codebase, captures useful project knowledge, and keeps verified decisions ready for the next session or model.

01 / Observe: Read the latest code, tools, files, and project state.
02 / Update: Save patterns and decisions confirmed through real work.
03 / Reuse: Make that context available to every new session and model.

Project context

Updating as Hena works

01
Observed from the codebase Catalog filters use URL state
Confirmed in use-product-filters.ts

Read
02
Verified through this task Product search follows the existing pattern
Search added · 42 tests passed

Saved
03
Ready for the next handoff Continue without another setup pass
Available to a new session or a different model

Ready

Context updated 3 verified facts

04 Model freedom

Switch models.
Keep the context.

Pick the model you want for any session, change it while you work, or let Hena route for quality. The project history, tools, and context stay attached.

75+: model providers
Any: OpenAI-compatible API
One: continuous project context

Incoming task Implement and test
the new billing flow

Model controlChoose or auto-route

Your call

Claude SonnetActive · coding + tools

Now

GPT CodexSwitch anytime · context ready

Next

Gemini ProSwitch anytime · context ready

Next

Connect the models you already use

OpenAIAnthropicGoogleOpenRouterBedrockAzureGitHub CopilotOllama + 67 more

05 Runtime choice

Your machine.
Or the cloud.

Choose the runtime that fits each project. Work locally when the code and tools should stay on your machine, or use a cloud workspace when the task needs to run remotely.

The context stays with the project.
Change runtime, session, or model without rebuilding what Hena already learned.

Free and MIT licensed · Source opening with the beta.

06 / Early access September 2026

All-in-one AI workspace

Bring all your AI work
into one place.

Join the waitlist for the first Hena desktop beta and open-source release.

Chat, code,automate.One living workspace.

Every kind of AI work.One workspace.

Friday admin, handled.

Different work.The same workspace.

Think in conversation.

Work inside the codebase.

Hand off repetitive web work.

Context that growswith the work.

Switch models.Keep the context.

Your machine.Or the cloud.

Bring all your AI workinto one place.

Every kind of AI work.
One workspace.

Different work.
The same workspace.

Context that grows
with the work.

Switch models.
Keep the context.

Your machine.
Or the cloud.

Bring all your AI work
into one place.