AI Providers
Roundtable supports five AI providers, giving you the flexibility to choose the models and pricing that best fit your team's needs. Each workspace connects to exactly one provider at a time, and you can switch providers at any point.
Supported Providers
Vertex AI
Vertex AI is Google Cloud's enterprise AI platform. Because Roundtable runs on GCP, Vertex AI is the most tightly integrated provider — no API key is required.
| Authentication | Automatic via Workload Identity (no API key) |
| Best for | Teams already on GCP who want zero-config AI |
| Supported models | Gemini 3.1 Pro Preview, Gemini 3.1 Flash-Lite, Gemini 2.5 Pro, Gemini 2.5 Flash, Gemini 2.0 Flash |
Vertex AI is the default provider for new workspaces. Since it uses Workload Identity, there's nothing to configure — it works out of the box.
Google AI Studio
Google AI Studio provides access to the same Gemini model family as Vertex AI, but authenticates with a personal API key instead of GCP Workload Identity. This makes it a great option for teams that aren't running on Google Cloud.
| Authentication | API key (from aistudio.google.com) |
| Best for | Non-GCP users who want access to Gemini models |
| Supported models | Gemini 3.1 Pro Preview, Gemini 3.1 Flash-Lite, Gemini 2.5 Pro, Gemini 2.5 Flash, Gemini 2.0 Flash |
OpenAI
OpenAI brings GPT-4 and the o-series reasoning models to Roundtable.
| Authentication | API key (from platform.openai.com) |
| Best for | Teams that prefer GPT models or need specific OpenAI capabilities |
| Supported models | GPT-4o, GPT-4o Mini, GPT-4.1, GPT-4.1 Mini, o3-mini |
Anthropic
Anthropic provides access to the Claude model family, known for strong reasoning and long-context performance.
| Authentication | API key (from console.anthropic.com) |
| Best for | Teams that prefer Claude's conversational style and safety features |
| Supported models | Claude Opus 4.5, Claude Sonnet 4.5, Claude 3.5 Haiku |
Ollama
Ollama lets you connect Roundtable to self-hosted open-source models. No API key is needed — just point Roundtable at your Ollama server.
| Authentication | None (self-hosted) |
| Best for | Air-gapped environments, cost control, or open-source model experimentation |
| Supported models | Llama 3.1 8B, Llama 3.1 70B, Qwen 2 7B, Mistral 7B, Code Llama 13B |
Ollama uses dynamic model discovery. Roundtable queries your Ollama server at connection time and lists every model you have pulled locally. If you pull a new model on your server, it will appear in Roundtable automatically the next time the model list is refreshed.
Provider Comparison
| Provider | Auth Method | API Key Required | Models | Strengths |
|---|---|---|---|---|
| Vertex AI | Workload Identity | No | Gemini 3.1 Pro Preview, 3.1 Flash-Lite, 2.5 Pro, 2.5 Flash, 2.0 Flash | Zero-config for GCP users, enterprise-grade |
| Google AI Studio | API Key | Yes | Gemini 3.1 Pro Preview, 3.1 Flash-Lite, 2.5 Pro, 2.5 Flash, 2.0 Flash | Same Gemini models without GCP dependency |
| OpenAI | API Key | Yes | GPT-4o, GPT-4o Mini, GPT-4.1, GPT-4.1 Mini, o3-mini | Industry-standard GPT models, reasoning with o3 |
| Anthropic | API Key | Yes | Claude Opus 4.5, Claude Sonnet 4.5, Claude 3.5 Haiku | Strong reasoning, large context windows |
| Ollama | None | No | Llama 3.1 8B/70B, Qwen 2 7B, Mistral 7B, Code Llama 13B + dynamic | Self-hosted, no per-token cost, fully private |
Setting API Keys
API keys are configured per-workspace through the workspace settings modal.
- Open your workspace.
- Click the Settings icon (gear) in the workspace header.
- Navigate to the AI Provider section.
- Select your provider from the dropdown.
- If the provider requires an API key, paste it into the API Key field.
- Click Save.
API keys are stored securely in GCP Secret Manager and are never exposed to other workspace members. Only workspace owners and admins can view or change the provider configuration.
Custom Model IDs
If a model you need isn't listed in the default dropdown, you can enter a custom model ID manually. This is useful when:
- A provider releases a new model that Roundtable hasn't added to its curated list yet.
- You're using a fine-tuned model with a custom identifier.
- Your Ollama server hosts a model variant with a non-standard name.
To use a custom model ID:
- Open Settings → AI Provider in your workspace.
- Select your provider.
- In the Model dropdown, choose Custom.
- Enter the full model identifier (e.g.,
ft:gpt-4o:my-org:my-finetune:abc123). - Click Save.
Roundtable passes the custom model ID directly to the provider's API. Make sure the ID is valid for your selected provider, or requests will fail.
Switching Providers on an Existing Workspace
You can change your workspace's AI provider at any time without losing conversation history or workspace data.
- Open Settings → AI Provider.
- Select the new provider from the dropdown.
- Enter an API key if required.
- Choose a model.
- Click Save.
The change takes effect immediately. New messages in the workspace will be handled by the updated provider and model. Previous conversation history is preserved, but keep in mind:
- Token counts may differ. Different providers tokenize text differently, so your token usage may change.
- Tool support is consistent. All 16 built-in tools work identically across every provider.
- System prompts carry over. Your workspace system prompt is provider-agnostic and does not need to be rewritten.