Platform
Everything ChatLite can do.
A light chat interface on top of a serious platform: model choice, research, creation, and the controls teams need to use AI safely.
Multi-model chat
One workspace for OpenAI, Anthropic Claude, Google Gemini, Mistral, DeepSeek, xAI Grok. Switch mid-conversation without losing context.
Deep research
Plan a search path, gather sources, compare claims, and return a cited brief.
Image generation
Generate images from seven providers in the same workspace as your text work.
Web search
Geo-aware, citation-backed answers grounded in live sources.
Guardrails
Screen every prompt and response for PII, secrets, injection, and policy breaches.
Code & data
Sandboxed Python, charts, Wolfram Alpha, and ~30 file formats.
Multi-model chat
Every leading model, one conversation.
ChatLite puts OpenAI, Anthropic Claude, Google Gemini, Mistral, DeepSeek, xAI Grok behind a single interface, including reasoning and multimodal models. Change model mid-thread and keep the whole conversation.
- Switch provider or model mid-conversation without losing context.
- Reasoning and multimodal models available in the same thread.
- Per-plan access, so teams control which models are in play.
- No separate tool, login, or billing relationship per vendor.
Deep research
Research that investigates a topic end-to-end.
Deep research is an agentic workflow rather than a single-shot answer. It plans a search path, gathers evidence from live web search and your uploaded files, compares claims across sources, and returns a structured, cited brief.
- Plans the search path before any searching begins.
- Gathers evidence from live web search and your uploaded files together.
- Compares claims across sources and tracks source quality.
- Returns findings, risks, and next steps with citation indices.
- Assign distinct models to the planner, researcher, verifier, and writer roles.
- Start, cancel, and watch runs as they progress.
Image generation
Generate images right where you chat.
One interface routes your prompt to any of seven image providers, so you can create, compare, and iterate without leaving the conversation.
- Configure size, style, format, and number of images per generation.
- Iterate on a prompt while keeping your full conversation history.
- Reference uploaded images to guide style and composition.
- Review image history and revisit earlier results.
- Usage tracked against your per-plan daily limits.
Web search
Answers grounded in the live web.
Built-in web search, powered by Brave Search, grounds answers in live sources rather than stale model memory. Every result carries citations your team can verify before sharing.
- Citations on every result, verifiable before anyone shares them.
- Geo-aware results, filterable by country, city, and region.
- Language-aware search that runs in the user's selected language.
- Toggle web search on or off per conversation.
- Only models that support web access are used for search.
Guardrails
Guardrails for governed AI.
Automated controls inspect and enforce your organization's policies on every prompt and every response, before content reaches a model or an end user.
- Detect and redact PII, secrets, and credentials in both directions.
- Block prompt injection and jailbreak attempts.
- Content policies tuned per team or department.
- A full audit trail of every guardrail decision.
Code & data
Run the analysis, not just the conversation.
ChatLite executes code in a sandbox, turns results into charts, and reads the file formats your work already lives in.
- Sandboxed Python execution with chart output.
- Wolfram Alpha for computation and structured knowledge.
- Around 30 file formats read for analysis.
- Connects to the tools on the integrations page.