Conflab logo Directed MULTI-PROVIDER Collaboration

Get more than one answer.

If you want the whole picture, one AI agent and one AI provider is not enough. Conflab is a directed multi-agent conversation workspace built for running multiple agents across multiple model providers like ChatGPT, Gemini, Claude, and more at the same time, so you can invent, explore, and learn through disagreement, comparison, and perspective instead of getting one polished answer from one system.

One agent is not enough If you want the whole picture, you need more than one answer. Conflab is built around comparison, pushback, and perspective.
Multiple providers, same room Run ChatGPT-style agents, Gemini-style agents, Claude-style agents, and others at the same time inside one directed workspace.
Discuss almost anything Science, politics, religion, medicine, black holes, multiple dimensions, strategy, writing, learning, or any other topic where a single AI answer leaves too much out.
Keep what matters Summaries, logs, and images live as persistent artifacts you can reopen later, embed elsewhere, or turn into something real instead of losing them in scrollback.
Lightning in a Bottle concept art showing a lightning-charged spherical structure in a storm
My House is on FIRE concept art showing a wildfire-defense roof sprinkler system protecting a house
I've Fallen and I Can't Get Up concept art showing an older adult doing balance movement practice in an autumn park
Lightning in a Bottle concept art showing aerial storm-energy craft over an island
I've Fallen and I Can't Get Up concept art comparing bone density and treatment imagery
My House is on FIRE concept art showing roofline sprinklers misting across a house at golden hour
Lightning in a Bottle concept art showing a lightning-charged spherical structure in a storm
My House is on FIRE concept art showing a wildfire-defense roof sprinkler system protecting a house
I've Fallen and I Can't Get Up concept art showing an older adult doing balance movement practice in an autumn park
Lightning in a Bottle concept art showing aerial storm-energy craft over an island
I've Fallen and I Can't Get Up concept art comparing bone density and treatment imagery
My House is on FIRE concept art showing roofline sprinklers misting across a house at golden hour
My House is on FIRE atmosphere image showing a house and pool with wildfire glow in the background
I've Fallen and I Can't Get Up concept art showing spine damage contrasted with recovery and treatment
Lightning in a Bottle concept art showing floating metallic pyramids suspended in a charged storm
My House is on FIRE concept art showing an aerial view of a sprinkler-protected house facing an approaching wildfire
I've Fallen and I Can't Get Up medical illustration showing jawbone deterioration under close examination
Lightning in a Bottle concept art showing a storm-lit saucer-shaped energy collector under lightning
Lightning in a Bottle concept art showing a close-up energy discharge arcing through a circular chamber
My House is on FIRE atmosphere image showing a house and pool with wildfire glow in the background
I've Fallen and I Can't Get Up concept art showing spine damage contrasted with recovery and treatment
Lightning in a Bottle concept art showing floating metallic pyramids suspended in a charged storm
My House is on FIRE concept art showing an aerial view of a sprinkler-protected house facing an approaching wildfire
I've Fallen and I Can't Get Up medical illustration showing jawbone deterioration under close examination
Lightning in a Bottle concept art showing a storm-lit saucer-shaped energy collector under lightning
Lightning in a Bottle concept art showing a close-up energy discharge arcing through a circular chamber
Persistent Drawers Settings, Conversations, Personas, and Results keep the app organized around real work instead of a single endless thread.
Conversation Control Parallel, sequential, and free-form modes let you choose whether the panel feels like a round table, a relay race, or a live argument.
Shareable Deliverables Saved summaries and logs become first-class artifacts you can reopen, export, print, embed, and build into something larger later.
Visual Follow-Through Summary-driven image generation turns the strongest parts of a conversation into a usable visual family rather than random decoration.
Free + Source Included Conflab is positioned as a totally free tool, with the source code included for people who want to inspect it, keep it, or build on it.

What It Is

The point of Conflab is not to pretend one model has many personalities. It is to give the Director a real orchestration surface for multiple voices across multiple providers at the same time, plus saved personas, reusable conversation setups, and durable output artifacts.

More than a chat window

Conflab is designed for debates, planning sessions, expert panels, writing rooms, invention sessions, and any situation where one response is not enough. You frame the objective, decide what kinds of minds should be in the room, and then work a panel that can mix models and providers at the same time instead of trapping the whole conversation inside one system.

That means you can use it for invention, science, politics, religion, medicine, strategy, education, writing, and other subjects where the whole value comes from hearing different lenses collide. If you want to explore black holes, multiple dimensions, treatment tradeoffs, or a product concept, different agents with different training can surface things one model by itself usually will not. You can put ChatGPT-style agents, Gemini-style agents, Claude-style agents, and other configured participants into the same directed workspace, stage disagreement on purpose, use saved personas as stable participants, keep current and historical conversations organized on disk, and end with outputs that survive after the session.

  • Assemble a panel with different roles, providers, models, or saved personas.
  • Use a totally free tool with the source code included.
  • Steer the discussion as Director instead of passively accepting one answer.
  • Compare how different providers reason, argue, and contribute inside the same conversation.
  • Generate summaries, logs, and image prompts from the strongest material.
  • Turn the conversation into something you can embed, present, or build into a product.
  • Return later and reopen the exact artifacts that mattered.
Conflab active conversation workspace
Live workspace with Director controls and multi-agent panel view. Main conversation surface
Conflab results drawer showing saved artifacts
Results drawer turns conversation outputs into persistent, browsable artifacts. Summaries, logs, images

Designed Around Outputs

Conflab’s workflow is visible in the interface itself: configure the environment, build the conversation, run the panel, then save what matters. The app is opinionated about keeping that chain intact.

Conflab image generation dialog from summary mode
Summary-driven image prompts let a conversation produce visual material that still belongs to the same thread of work. Image workflow
Conflab summary viewer modal
Saved deliverables open in their own viewer, ready to export, print, or share. Artifact viewer

Conversation Showcase

These are real sample artifacts exported from the local workspace. Each card links to a copied HTML summary and log inside this site folder, and the summaries use the finished exported versions with their images already embedded.

Lightning in a Bottle sample image showing lightning capture concept Lightning in a Bottle secondary sample image
11 rounds Speculative engineering Summary + log

Lightning in a Bottle

A stress test of an atmospheric-electricity concept that started as “is this impossible?” and ended with concrete technical constraints, a prototype pathway, and a business model narrow enough to be believable.

My House is on FIRE! sample image showing wildfire defense system in action My House is on FIRE! atmosphere image
8 rounds Home defense planning Summary + log

My House is on FIRE!

A practical, emotionally charged conversation about wildfire defense, deployable roof systems, aesthetics, cost pressure, and the difference between an engineerable solution and a brochure fantasy.

The Workflow Shape

The app already has a clear product story hiding inside the interface. This section makes that shape explicit so the site can explain Conflab without drowning people in feature lists.

01

Set the environment

Choose a root folder, connect your model providers, pick a helper model, and give Conflab a stable place to save the work.

02

Build the room

Create a conversation, define the objective, describe the kinds of participants you want, and assign the panel.

03

Direct the panel

Push the room in parallel, sequential, or free-form mode. Ask for disagreement, synthesis, pressure tests, or deliverables.

04

Keep the artifacts

Save summaries, logs, and images so the conversation becomes a reusable asset rather than a moment that disappears.

From rough question to saved deliverable.

This first-pass local site is intentionally simple under the hood: one HTML page, external assets, and portable example artifacts. That gives you a clean base for GitHub Pages or a real `conflab.ai` deployment later, while still being easy to edit locally right now.

Coming soon.