Panel creator greentfrapp claims the released research workspace lets Claude Code create custom viewers alongside shared files, PDFs, and executable notebooks, reducing interface switching and bespoke visualization work during agent-assisted research.

state: corroboratedheat: lowuncertainty: mediumconvergesscott: mediumresearch-agents agent-workspaces knowledge-systemsgreentfrappPanel

What is this?

Panel is an early MIT-licensed open-source research workspace by greentfrapp (Show HN 2026-09-15; runnable repo fetched first-hand per the case; no further background on the creator appears in the supplied material): dockable panes over a live Jupyter kernel where Claude Code — the only fully supported agent — can author its own custom viewer panes alongside shared files, PDFs, and executable notebooks, with persistent workspaces and shared human+agent notebook editing; its claimed benefit of less interface switching and less bespoke visualization work is creator-stated and unverified. The fresh web pass returned no direct hits on Panel — no adoption data, review, or independent test appears in the results. What it does show is the surrounding territory being built out from the vendor side: Anthropic is embedding interactive apps (Slack, Figma, Asana) inside Claude's interface and pushing Cowork as a shared asynchronous workspace explicitly pitched on avoiding window switches, while Google and Microsoft ship notebook-style AI research workspaces (Gemini Notebooks, Copilot Notebooks) and small third-party builds like aTerm standardize multi-pane agent+shell layouts. Nothing here changes the case's standing read: a corroborated-at-pattern-level but unproven-at-product-level early build, now sitting in a space incumbents are visibly moving into.

Why it matters to Scott

Holds as converges, with a sharper edge from the fresh pass: Panel (and sibling Radix) independently build the boundary-generated-interface claim Scott's Agent-Native Computing and Perceptual Engineering already hold, while the new web material shows incumbents (Anthropic's embedded apps and Cowork, Gemini/Copilot Notebooks) arriving at the workspace-over-chat frame through fixed human-authored surfaces — validating demand yet leaving agent-authored panes to small builders, which is both a dated-receipts contrast to publish against the framework and a runnable design-space reference for dev:project.aidiscoveryui. Medium rather than high: the products remain early, unverified, and quietly attended, so this is reference material and a sharpened contrast, not anything that changes what Scott builds or argues.

ip:framework.agent-native-computingip:concept.perceptual-engineeringdev:project.aidiscoveryuidev:concept.synchronized-conversational-visual-interfaceradar:concept.generative-uiradar:concept.research-agentsradar:concept.claude-coderadar:hearth-shared-context-agent-apps

queries asked of Scott's wikis

boundary-generated interfaces agents building their own UI
perceptual engineering agent interface design
aidiscoveryui design references agent-authored interfaces
shared human-agent workspaces notebook co-editing
single-agent harness lock-in Claude Code coupling portability
research agent knowledge workspace literature review design

Measured heat

now 0 pts/hpeak 0 pts/hcomments 0/hpeers p14momentum: steady2 platformsage 626h

points/hour across evidence · reading as of 2026-10-12 02:59:37.977291+11:00 · deterministic, not a model opinion

How the heat travelled

09-15 14:26 (minted)	⭐ origin echo-reconstructed	Panel is an early build for testers with persistent workspaces, shared notebook editing, background commands, literature reviews, and custom greentfrapp on github (echo) · attributed from hn.story.49712621 · published time unknown	—
09-15 13:58	first on hacker news · published · lag ?	Show HN: Panel – A research workspace where the agent can build its own panes greentfrapp	—
09-15 13:58	amplified on hacker news 👑	hn.story.49712621 greentfrapp	peak 54 · 23 comments · 70% of case engagement
09-24 18:35	amplified on hacker news	hn.story.49834964 0x1062	peak 15 · 18 comments · 30% of case engagement
09-15 14:21	our radar first saw it · lag ?	discovery anchor: hn.story.49712621	—

pace: p68 vs 1032 stories at the 336h mark (now 626h old) — ahead of gemini-agentic-video-understanding (1.0x), behind claude-cowork-windows-update-command-failure (1.0x)

Evidence (3) — ⭐ canonical anchor

source	object	author	score	comments
🟧 hn	Show HN: Panel – A research workspace where the agent can build its own panes Retrieved article excerpt Open article · Retrieved 2026-09-15T14:22:28.968326+00:00 # Panel A research workspace where the agent works beside you: chat, files, PDFs and notebooks in one dock, and the agent can also create custom viewers and apps when necessary This is an early build for testers. Expect rough edges, and feel free to raise issues. [Screenshot of Panel in action.](https://github.com/greentfrapp/panel/blob/main/assets/panel_screenshot.jpg) ## Before you start - [Node](https://nodejs.org/en/download) 22.18 or newer (or 24.12 and newer) - [pnpm](https://pnpm.io/installation) - [uv](https://docs.astral.sh/uv/getting-started/installation/), which fetches the Python it needs (3.12 or newer) by itself - [Claude Code](https://code.claude.com/docs/en/setup), installed and signed in: run `claude` once and log in. The agent and the literature review run through it. ## Install and start ``` pnpm install uv sync pnpm start ``` Then open <http://localhost:4173>. `pnpm start` builds the app first, so the first start takes a minute. Ctrl-C stops everything it started. ## Where your things are - `~/Panel/panel.db` holds your conversations and everything the agents did. - `~/Panel/workspaces` is where new Workspaces are created, unless you pick another folder. Both are outside this folder, so deleting or re-cloning the repo keeps them. ## What works - Chatting with an agent that can read and write files, and asks before running a tool. - Workspaces: a folder the agent works in, with its own chats and saved layout. - Panes for files, PDFs, markdown and Jupyter notebooks. Notebooks run against a real kernel, and you and the agent can edit the same one. - Long-running commands in the background, which you can watch and stop. - Panes the agent writes for you when you ask to see something a built-in Pane cannot show. - A literature review: ask the chat for one, and open its result from the tool card. ## What doesn't yet - Currently only has full support for Claude Code. - Modules start only by asking the chat. There is no button to launch one. - The hypothesis Modules have no view of their own, so their results can be hard to read. - Modules don't work with OpenAI API yet. ## The OpenAI key (optional) Copy `apps/server/.env.example` to `apps/server/.env` and set `OPENAI_API_KEY`. This adds "OpenAI API" to the agent picker, for chat and tools. It does not run literature reviews or the hypothesis Modules: those need an agent that can search the web, and today only Claude Code can. Without a key, the picker shows OpenAI as not set up, which is expected. ## If something's wrong - "Panel couldn't reach its server." The server half is not running. Check the terminal `pnpm start` is in, then press Retry. - An agent shows as not set up. The reason is written under the message box. - A port is already in use, or the app answers but never loads: run `pnpm dev:doctor`. It says what is holding each port and how to clear it. ## Licence [MIT](https://github.com/greentfrapp/panel/blob/main/LICENSE) --- ## The idea ### UI The UI has multiple configurable windows, called Panes, that can display things ranging from image files, data files, code, as well as chat sessions. This is critical for researchers who often have to context switch between different types of files. A default set of Panes are provided for common use cases. But custom Panes can also be added by humans and agents, such as a PDB viewer or SQLite visualizer. ### Module Protocol Modules are similar to Skills but with additional definitions to support inter-module workflows and integration with the workspace. Specifically, Modules have typed definitions for Inputs, Outputs, and Intermediates. Inputs and Outputs are straightforward. Intermediates refer to objects that provide observability, such as the Chain-of-Thought or scratchpad for an agentic Module, or may be intermediate outputs in a multi-stage Module. These are especially important for processes that need transparency or long-running jobs that should show progress. Having typed definitions for these enable validation at runtime and make it easier for humans and agents to develop custom Modules for downstream tasks and Panes for visualizations. #### Data Abstraction Layer A data abstraction layer (DAL) bridges in-memory and filesystem objects. A DAL helps to map a URI to either an in-memory store or a local file, so that the Module just has to concern itself with the manipulation of the object.	greentfrapp	54	21
🟧 echo.github ⭐	Panel is an early build for testers with persistent workspaces, shared notebook editing, background commands, literature reviews, and custom	greentfrapp	—	—
🟧 hn	Show HN: Radix – Visual UI for agentic programming	0x1062	15	18

source

object

author

score

comments

🟧 hn

Show HN: Panel – A research workspace where the agent can build its own panes

Retrieved article excerpt

Open article · Retrieved 2026-09-15T14:22:28.968326+00:00

# Panel

A research workspace where the agent works beside you: chat, files, PDFs and notebooks in one dock, and the agent can also create custom viewers and apps when necessary

This is an early build for testers. Expect rough edges, and feel free to raise issues.

[Screenshot of Panel in action.](https://github.com/greentfrapp/panel/blob/main/assets/panel_screenshot.jpg)

## Before you start

- [Node](https://nodejs.org/en/download) 22.18 or newer (or 24.12 and newer)
- [pnpm](https://pnpm.io/installation)
- [uv](https://docs.astral.sh/uv/getting-started/installation/), which fetches the Python it needs (3.12 or newer) by itself
- [Claude Code](https://code.claude.com/docs/en/setup), installed and signed in: run `claude` once and log in. The agent and the literature review run through it.

## Install and start

```
pnpm install
uv sync
pnpm start
```

Then open <http://localhost:4173>. `pnpm start` builds the app first, so the first start takes a minute. Ctrl-C stops everything it started.

## Where your things are

- `~/Panel/panel.db` holds your conversations and everything the agents did.
- `~/Panel/workspaces` is where new Workspaces are created, unless you pick another folder.

Both are outside this folder, so deleting or re-cloning the repo keeps them.

## What works

- Chatting with an agent that can read and write files, and asks before running a tool.
- Workspaces: a folder the agent works in, with its own chats and saved layout.
- Panes for files, PDFs, markdown and Jupyter notebooks. Notebooks run against a real kernel, and you and the agent can edit the same one.
- Long-running commands in the background, which you can watch and stop.
- Panes the agent writes for you when you ask to see something a built-in Pane cannot show.
- A literature review: ask the chat for one, and open its result from the tool card.

## What doesn't yet

- Currently only has full support for Claude Code.
- Modules start only by asking the chat. There is no button to launch one.
- The hypothesis Modules have no view of their own, so their results can be hard to read.
- Modules don't work with OpenAI API yet.

## The OpenAI key (optional)

Copy `apps/server/.env.example` to `apps/server/.env` and set `OPENAI_API_KEY`. This adds "OpenAI API" to the agent picker, for chat and tools.

It does not run literature reviews or the hypothesis Modules: those need an agent that can search the web, and today only Claude Code can. Without a key, the picker shows OpenAI as not set up, which is expected.

## If something's wrong

- **"Panel couldn't reach its server."** The server half is not running. Check the terminal `pnpm start` is in, then press Retry.
- **An agent shows as not set up.** The reason is written under the message box.
- **A port is already in use**, or the app answers but never loads: run `pnpm dev:doctor`. It says what is holding each port and how to clear it.

## Licence

[MIT](https://github.com/greentfrapp/panel/blob/main/LICENSE)

---

## The idea

### UI

The UI has multiple configurable windows, called Panes, that can display things ranging from image files, data files, code, as well as chat sessions. This is critical for researchers who often have to context switch between different types of files.

A default set of Panes are provided for common use cases. But custom Panes can also be added by humans and agents, such as a PDB viewer or SQLite visualizer.

### Module Protocol

Modules are similar to Skills but with additional definitions to support inter-module workflows and integration with the workspace.

Specifically, Modules have typed definitions for Inputs, Outputs, and Intermediates.

Inputs and Outputs are straightforward. Intermediates refer to objects that provide observability, such as the Chain-of-Thought or scratchpad for an agentic Module, or may be intermediate outputs in a multi-stage Module. These are especially important for processes that need transparency or long-running jobs that should show progress.

Having typed definitions for these enable validation at runtime and make it easier for humans and agents to develop custom Modules for downstream tasks and Panes for visualizations.

#### Data Abstraction Layer

A data abstraction layer (DAL) bridges in-memory and filesystem objects. A DAL helps to map a URI to either an in-memory store or a local file, so that the Module just has to concern itself with the manipulation of the object.

greentfrapp

🟧 echo.github ⭐

Panel is an early build for testers with persistent workspaces, shared notebook editing, background commands, literature reviews, and custom

greentfrapp

—

🟧 hn

Show HN: Radix – Visual UI for agentic programming

0x1062

Interpretation history

2026-09-24T22:14:00Z
grounded: converges/medium — Holds as converges, with a sharper edge from the fresh pass: Panel (and sibling Radix) independently build the boundary-generated-interface claim Scott's Agent-
2026-09-24T22:06:54Z
Radix — an independent Show HN product where the agent generates persistent local interactive widgets — plus a third hobbyist in the Panel thread describing model-mounted custom panels, lifts this from a solo Panel episode to a corroborated pattern: agent-authored interfaces now have multiple independent implementations, though neither product's claimed efficiency gains are verified.
2026-09-24T20:37:41Z
evidence attached: hn.story.49834964 — Independent second product with the same thesis — agent-authored persistent interactive widgets replacing chat — corroborating spread of the pattern beyond Panel.
2026-09-15T14:29:07Z
grounded: converges/medium — Panel’s claimed agent-built viewers converge with Scott’s Agent-Native Computing boundary-generated interfaces and Perceptual Engineering, while offering a conc
2026-09-15T14:26:25Z
case created — The runnable early release establishes a distinct agent-authored interface episode rather than evidence for a broad research-workspace adoption trend.

Decision trace

10-07 19:40review_dormantscheduled targets exhausted or 28 quiet days
10-07 19:40drop_targetsquiet through full ladder or over cap 8
09-25 08:14repriceRadix — an independent Show HN product where the agent generates persistent local interactive widgets — plus a third hobbyist in the Panel thread describing model-mounted custom panels, lifts this fro
09-25 08:14groundHolds as converges, with a sharper edge from the fresh pass: Panel (and sibling Radix) independently build the boundary-generated-interface claim Scott's Agent-Native Computing and Perceptual Eng
09-25 06:37attachIndependent second product with the same thesis — agent-authored persistent interactive widgets replacing chat — corroborating spread of the pattern beyond Panel.
09-25 06:34propose_attachIndependent second product with the same thesis — agent-authored persistent interactive widgets replacing chat — corroborating spread of the pattern beyond Panel.
09-19 03:56review_screenThe change adds only a feature question about checkpoints and removes a question about preserving agent-created layouts; it provides no new implementation result, release detail, or credible evidence
09-16 06:27review_screenThe added comments reiterate the already established Panel concept—an agent-assisted workspace with shared research panes and agent-generated interfaces—without providing a new implementation result,
09-16 00:29groundPanel’s claimed agent-built viewers converge with Scott’s Agent-Native Computing boundary-generated interfaces and Perceptual Engineering, while offering a concrete extension beyond the model-populate
09-16 00:26createThe runnable early release establishes a distinct agent-authored interface episode rather than evidence for a broad research-workspace adoption trend.