The most feature-rich open source AI coding agent.
110+ AI providers, multi-agent orchestration, cross-session memory, and everything you need in one agent.
AgentiCode combines the best ideas from every major AI coding agent into a single, provider-agnostic platform:
| From | What We Took |
|---|---|
| Core Engine | Full agent engine — bridge, swarm, buddy, memory, vim, 40+ tools, CLI |
| OpenCode | 110+ provider registry via models.dev, Electron desktop concepts |
| Aider | Repo map, auto-lint/test loop, git auto-commit, voice coding |
| Cline | Checkpoint/restore, screenshot input, browser automation |
| Cursor | @-mentions context system (@file, @tree, @git, @search) |
| Goose | Recipe YAML workflows, sandbox containers, auto-compaction, skills |
| Claurst | Conversation fork, session budget, advisor dual-model review |
| ClawCode | Experience learning (ECAP), multi-provider abstraction pattern |
| Category | What's Included |
|---|---|
| Providers | 110+ via models.dev — OpenAI, Anthropic, Google, Groq, DeepSeek, Mistral, xAI, OpenRouter, Ollama, LM Studio, and 100 more. Online + local. |
| Multi-Agent | Swarm orchestration, coordinator mode, parallel workers, team management, permission bridge, advisor review |
| Memory | Cross-session persistence (4 types: user/feedback/project/reference), experience learning (ECAP), project hints (.agenticode/hints.md) |
| Context | @-mentions (@file, @tree, @git, @url, @search), image/screenshot input, repo map code graph |
| Remote | Bridge sessions (WebSocket + hybrid transport), remote REPL, session sync, reconnection with backoff |
| Workflows | Recipe YAML system with parameters and sub-recipes, planner model separation |
| Safety | PowerShell destructive command detection, sandbox container isolation, advisor dual-model review |
| Dev Tools | Auto-lint/test loops, git auto-commit, checkpoint/restore snapshots, smart auto-compaction |
| UX | Buddy companion sprites (18 species), vim mode, conversation fork/branch, session budget tracking |
| Export | Session replay in Markdown, JSON, or HTML. Full conversation history with cost tracking. |
| Browser | Playwright/Puppeteer MCP integration with presets for headless testing |
| Search | Web search via Tavily, Brave, SearXNG, or custom backends |
git clone https://github.com/kyaky/agenticode.git
cd agenticode
bun install
bun startSet any provider API key:
export OPENAI_API_KEY=sk-... # OpenAI
export GROQ_API_KEY=gsk_... # Groq
export DEEPSEEK_API_KEY=sk-... # DeepSeek
export AGENTICODE_API_KEY=sk-ant-... # Anthropic
# ... or any of 110+ providerssrc/
├── bridge/ # Remote sessions (WebSocket + hybrid transport)
├── buddy/ # AI companion sprites (18 species, 5 rarities)
├── coordinator/ # Multi-agent orchestration + advisor review
├── memdir/ # Cross-session memory + ECAP experience learning
├── recipe/ # YAML workflow engine
├── vim/ # Modal editing state machine
├── tools/ # 40+ tools + auto-fix + auto-commit + repo-map
├── services/
│ ├── api/ # Core query engine
│ └── providers/ # Provider abstraction layer
│ ├── types.ts # Unified BaseProvider interface
│ ├── openai.ts # OpenAI + 100 compatible endpoints
│ ├── anthropic.ts # Native Anthropic HTTP client
│ ├── models-registry.ts # models.dev 110+ providers
│ ├── web-search.ts # Tavily/Brave/SearXNG
│ ├── capabilities.ts # Thinking/caching/tool detection
│ ├── converter.ts # Format: Internal↔OpenAI↔Anthropic↔Gemini
│ └── router.ts # Auto-detection + factory
├── utils/
│ ├── mentions.ts # @-mentions context system
│ ├── imageInput.ts # Screenshot/image input
│ ├── browserAutomation.ts # Playwright/Puppeteer MCP
│ ├── sessionExport.ts # Session replay (MD/JSON/HTML)
│ ├── sandbox.ts # Docker container isolation
│ ├── checkpoint.ts # File snapshot/restore
│ ├── budget.ts # Token/cost limits
│ ├── fork.ts # Conversation branching
│ └── auto-compact.ts # Smart context management
├── cli/ # Terminal UI + 87 commands
└── components/ # Ink React terminal components
AgentiCode integrates with models.dev for automatic provider and model discovery:
import { searchModels, createProviderFromRegistry } from './services/providers'
// Search across all providers
const models = await searchModels("gpt-4")
// Auto-create provider from registry
const provider = await createProviderFromRegistry("groq", "llama-3.3-70b-versatile")
// Find cheapest model with tool calling
const cheap = await findCheapestModel({ needsToolCall: true })Online providers: OpenAI, Anthropic, Google, Groq, DeepSeek, Mistral, xAI, OpenRouter, Together, Fireworks, Cerebras, Perplexity, Cohere, 302.AI, Alibaba, GitHub Copilot, GitHub Models, Azure, Bedrock, Vertex, and 90+ more.
Local providers: Ollama, LM Studio, llama.cpp, Jan, vLLM, LocalAI.
MIT