Big-AGI is a chat workspace that lets you use several AI models — OpenAI, Anthropic, Google Gemini and more — from a single screen. It is developed by Token Fabrics in the United States and has been maintained as open source (MIT license) since its first release in March 2023. Its standout feature is “Beam & Merge”: send the same question to multiple models at once, line the answers up side by side, and combine the best parts into a single response. You bring your own API keys and connect directly to each provider (BYO), and conversation data is stored on your own device by default.
Key Features
- Beam & Merge (parallel runs and consolidation): Send one prompt to multiple models simultaneously and compare the replies side by side. The official site mentions running up to 24 models in parallel, which makes inconsistencies and mistakes easier to spot. You can also merge the good parts of several answers into one
- Broad provider support: Alongside cloud providers such as OpenAI, Anthropic, Google Gemini, Mistral, Groq, DeepSeek, xAI, Perplexity, AWS Bedrock and Azure, it also works with local runtimes like Ollama, LocalAI and LM Studio. Any OpenAI-compatible endpoint can be added with automatic detection
- Local-first design: Conversations and personas are saved on your device by default. Cloud backup and multi-device sync are positioned as paid-plan options, so you decide where your data lives
- Transparent requests: The AI Inspector shows the exact API request that was sent, the tokens used and the estimated cost, making it easy to see how much you are spending on each model
- Personas and fine-grained parameters: Assign documents and custom instructions to task-specific personas and switch between them. Many parameters, including temperature and reasoning depth, can be tuned individually
- Self-hostable: Supports Docker, one-click deployment to Vercel, and local execution with Node.js, so running it on your own server is an option
Pricing
API usage is billed directly by each AI provider under the bring-your-own-key model, and the official site states there is no markup from Big-AGI. The figures below are for the workspace itself.
| Plan | Price | Main features |
|---|---|---|
| Free | $0 | The full open-source feature set, multi-model features, personas, hosting by Big-AGI |
| Pro | $9/month ($108/year) | Everything in Free, plus cloud backup of chats and personas, multi-device sync, email support, priority access to new features |
| Business | Contact for pricing | Custom deployments, team management, SLA |
Pricing is as of August 2026. Please check the official site for the latest pricing.
Pros & Cons
✅ Pros
- Seeing several models’ answers to the same question side by side makes contradictions and errors easier to catch
- No need to subscribe to each model’s own service separately — you switch between them on one screen
- Open source under the MIT license, so you can self-host and keep it under your own control
- Because the API keys are yours, you pay each provider’s usage rates directly with no middleman margin
- Local-first design lets you choose where conversation data is stored
⚠️ Cons
- You have to obtain and configure API keys yourself, which is a hurdle for people new to AI services
- Unlike a flat monthly chat service, you pay for what you use, so costs are harder to predict
- Running several models in parallel multiplies API charges accordingly
- The interface is dense with information and may feel excessive if you only want a simple chat
- Syncing conversations across multiple devices requires the Pro plan
Comparison with Similar Services
| Item | Big-AGI | LibreChat | Open WebUI | TypingMind |
|---|---|---|---|---|
| License | Open source (MIT) | Open source | Open source | Commercial (one-time / subscription) |
| Main strength | Parallel model runs and answer consolidation | Feature-rich multi-user operation | Strong fit with local LLMs (Ollama) | Ease of use and polished UI |
| Self-hosting | Yes (Docker / Vercel) | Yes | Yes | Partially |
| API keys | Bring your own | Bring your own | Bring your own | Bring your own |
| Hosted version | Yes (Free / Pro) | Mostly self-run | Mostly self-run | Yes |
Who Is It For
- People who want to compare answers from multiple AI models and judge for themselves which one holds up
- Anyone doing important research who would rather not take a single model’s answer at face value
- Those who already hold API keys from several providers and want to use them without adding more subscriptions
- People who want conversation data under their own control, or who want to run the workspace on a company server
- Users who want to mix local LLMs (such as Ollama) with cloud models on the same screen
Summary
Big-AGI is a workspace built around one strength: running several AI models side by side and cross-checking their answers. It takes more setup than a single chat service, but since you hold the API keys yourself there is no extra middleman cost, and being open source means you can also run it on your own server. Start with the hosted free plan to see how Beam & Merge feels in practice, and consider Pro once you need to sync across multiple devices.