Mysti – An AI‑powered coding teammate for VS Code

What it is – Mysti is a Visual Studio Code extension that lets you talk to, switch between, and combine dozens of code‑generation AIs inside the editor. It supports 12 providers (Claude Code, Codex, Gemini, GitHub Copilot, Cline, Cursor, OpenClaw, Manus, OpenCode, Qwen Code, Ollama, LocalAI) and adds a set of collaboration‑oriented features such as multi‑agent brainstorming, persona selection, autonomous execution, and fine‑grained permission controls.

---

Core ideas

Idea	Why it matters
Multi‑provider	You can keep the subscriptions you already have and avoid lock‑in. Switching providers is a single click.
Brainstorm mode	Two agents discuss a problem using one of five strategies (quick, debate, red‑team, perspectives, Delphi). The extension detects convergence and can stop the discussion early.
Personas & skills	16 built‑in developer personas (architect, debugger, security‑focused, etc.) and 12 toggleable skills (concise, test‑driven, auto‑commit, …) shape the tone and behaviour of the AI.
Autonomous mode	The AI can act on files without further prompts, subject to three safety levels and a permission model (read‑only, ask‑permission, full‑access).
@‑mentions	Inline tags like @claude or @src/auth.ts route a request to a specific agent and/or give it temporary file context.
Context compaction	When token usage approaches the model’s limit, Mysti automatically summarizes earlier messages to keep the conversation within bounds.

---

Key features (as described in the README)

• 12 AI providers – cloud (Claude, Codex, Gemini, Copilot, etc.) and local (Ollama, LocalAI). |
• Brainstorm mode – pick any two agents, choose a collaboration strategy, and let them debate or co‑create. |
• 16 developer personas – change the AI’s perspective with a single click or via the toolbar. |
• Autonomous execution – safety‑classifier + three safety modes, learning‑memory of your permission preferences, audit‑trail logging. |
• @‑mention routing – direct a request to a specific agent or file, chain responses, and keep context local to the panel. |
• Context compaction – automatic summarisation when token usage > 75 % of the model window. |
• Conversation history – all chats are persisted and searchable inside VS Code. |
• Slash‑command palette – quick access to skills, personas, and provider‑specific actions. |
• Extensive settings – provider paths, brainstorm agents/strategy, autonomous safety mode, token budgets, compaction thresholds, etc. |
• Keyboard shortcuts – Ctrl/Cmd+Shift+M to open Mysti, Ctrl/Cmd+Shift+N for a new tab. |

---

Installation & quick start (from the README)

1. Install the extension
   • In VS Code press Ctrl+P (or Cmd+P), type ext install DeepMyst.mysti and hit Enter, or click the Install from VS Code Marketplace badge in the README.
2. Install at least one provider CLI (the extension talks to the provider through its command‑line client). Example for Claude Code:

   npm install -g @anthropic-ai/claude-code
   claude auth login
   

   Other supported CLIs include GitHub Copilot, Gemini, Codex, Cline, Cursor, OpenClaw, OpenCode, Qwen Code, Ollama, and LocalAI. Install any two if you want to use Brainstorm mode.
3. Open Mysti – click the Mysti icon in the Activity Bar or press Ctrl+Shift+M (Cmd+Shift+M).
4. Start coding – type a natural‑language request (e.g., “Add unit tests for auth.ts”) and let the AI respond. Use @ mentions, slash commands, or the persona toolbar to refine the interaction.

---

Configuration highlights (excerpt from the README)

{
  "mysti.defaultProvider": "claude-code",
  "mysti.brainstorm.agents": ["claude-code", "google-gemini"],
  "mysti.brainstorm.strategy": "quick",
  "mysti.accessLevel": "ask-permission",
  "mysti.autonomous.safetyMode": "balanced",
  "mysti.compaction.enabled": true,
  "mysti.compaction.threshold": 75
}

• Provider‑specific paths (mysti.claudePath, mysti.ollamaPath, …) let you point to custom installations.
• Brainstorm settings control which two agents collaborate, the chosen strategy, auto‑convergence, and max discussion rounds.
• Autonomous settings let you pick a safety mode and define block patterns.
• Compaction settings toggle summarisation and set the token‑usage threshold.

---

Requirements (as listed)

CLI	Subscription needed?	Install command
Claude Code	Anthropic API / Claude Pro/Max	npm install -g @anthropic-ai/claude-code
GitHub Copilot	Copilot Pro/Business	npm install -g @github/copilot-cli
Gemini	Google AI API	npm install -g @google/gemini-cli
Codex	OpenAI API	follow OpenAI guide
Cline	Depends on model provider	npm install -g cline
Cursor	Cursor subscription	`curl https://cursor.com/install -fsS
OpenClaw	OpenClaw account	npm install -g openclaw@latest && openclaw onboard --install-daemon
OpenCode	Any of Anthropic, OpenAI, Google, Groq keys	npm i -g opencode-ai@latest
Qwen Code	Alibaba Qwen API	npm install -g @qwen-code/qwen-code@latest
Ollama	Local – no subscription	download from ollama.com
LocalAI	Local – no subscription	download from localai.io

Only one CLI is required to start; install a second one to enable brainstorming.

License & community

• License: Apache 2.0 (as shown by the badge).
• Marketplace stats: version badge, install count, rating, stars, forks are displayed in the README, indicating an active public release.
• Documentation: extensive markdown docs (Providers, Brainstorm, Personas, Autonomous Mode, Mentions, Compaction, Architecture, Features) are linked from the README.

---

TL;DR

Mysti turns VS Code into a collaborative AI coding hub. By wiring together up to twelve different code‑generation models, it lets you switch, combine, or run them autonomously, all while giving you granular control over safety, permissions, and context. Install the extension, add the CLI for the providers you already pay for (or run a local model), and start chatting with your new AI coding team.