Dolphin Pod is a modern, light & fast group chat-like app where the members are AI agents. Roger moderates, Lincoln orchestrates,
and a team of specialists (coding, long-horizon projects, security, debugging, writing, docs, images and
teaching, fun facts, verification, web research and git) answer in shared chats, DP Council gives big tasks one sharp critique first, and in the development chats Vera directs engineers, designers and game specialists. Every agent runs on the local or cloud reasoning model you choose.
Highlights
- Chats with a crew: Coding, General, Long Horizon Agentic, Learning, Creative and the Software, Web and Game Development chats out of the box, plus your own.
- Routing that thinks: Roger acknowledges instantly, decides intent and mode (learn, quick, plan, implement),
Lincoln briefs the right specialist, and
@mentions go straight to an agent.
- Plans before power: planning runs read-only; approving a plan (button or "approve") grants full
agentic permissions inside a workspace folder. Every action is logged.
- Any provider: Ollama, LM Studio, custom OpenAI-compatible servers, OpenRouter, OpenAI, Anthropic,
Google Gemini, xAI, Mistral, DeepSeek, Groq, Together, Fireworks and Azure OpenAI. Only reasoning models
are selectable, and the effort control adapts to each model.
- Streamed thinking from every provider, full Markdown with code highlighting and math.
- Smart context handling: conversations are compacted automatically before they fill a model's context window
(older messages become a saved summary, never silently dropped), large tool output (logs, files, search hits,
JSON) is compressed by content type with the full text one tool call away, and prompts are laid out for provider
prompt caching. Only the relevant parts of your custom instructions are sent.
- Sync and iPhone: optional end-to-end encrypted sync (Postgres/NeonDB, your passphrase, a recovery key) keeps
chats, agents, settings and provider keys identical across desktops and the iPhone app, and the iPhone can start
work on a desktop, nearby over Wi-Fi or through the cloud.
- Housekeeping: rolling 24-hour cleanup per chat, automations at a time of day, and Roger can change any
setting from chat.
- Frosted glass UI in dark and light, personalizable agent colors and gradients, hover cards with all-time
usage and timing.
Development
Requirements: Rust (stable), Node 22+, pnpm 10, and the Tauri prerequisites.
pnpm install
pnpm dev
pnpm dev starts the desktop app with hot reload. To work on the UI in a plain browser (with a mock backend), run
pnpm --filter @dolphin/desktop dev and open http://localhost:1420.
| Task | Command |
|---|
| Core tests | cargo test -p dolphin-core |
| Frontend tests | pnpm --filter @dolphin/desktop test |
| Regenerate TS types from Rust | pnpm bindings |
| Regenerate all icons | python3 packages/brand/scripts/make_icons.py |
| Live run against Ollama | cargo run -p dolphin-core --example live_ollama -- qwen3.5:2b |
| Release build | pnpm --filter @dolphin/desktop tauri build |
| iOS framework and app | see apps/ios |
| Website dev server | pnpm --filter @dolphin/website dev |
| Website production build | pnpm --filter @dolphin/website build |
Data lives in the OS app-data folder (dolphin-pod.db, SQLite). API keys are stored in the system keychain.
Credits
Built by MintFlow Technologies.
Lead developer: Luca Montanari (@montanarii.luca).
© 2026 MintFlow Technologies. All rights reserved. See LICENSE.
