Recipes
Recipes
Copy-paste solutions grouped by theme. Every recipe end-to-end, runs as written.
60+ recipes. Each one: single page, install line, code, run. No dependency between recipes — every page stands alone.
#Popular outcomes
- Support agent — pair with Persistent memory, Confirmation-gated tool, and Audit log
- Research agent — pair with Research team, Background agents, and Eval suite
- Code agent — pair with Code reviewer, Mandatory sandbox, and Deterministic replay
- Internal copilot — pair with Doc loaders, RAG chat, and Prompt injection
#Getting started
- Custom adapter — wrap any LLM API
- More providers — OpenAI, Anthropic, Gemini, local
- Simulate stream — fake streaming for one-shot providers
- Framework adapters — React / Vue / Svelte / Solid / Ink / RN / Angular
#Chat UI
- RAG chat · PDF Q&A
- Cost-guarded chat · Persistent memory
- Edit + regenerate · Progressive tool calls
- Multi-modal · Generative UI
#RAG + retrieval
#Memory
- Persistent memory · Hierarchical memory
- Auto-summarize · Virtualized memory
- Graph memory · Personalization
- Encrypted memory
#Adapter composition
#Tools + integrations
#Multi-agent + orchestration
#Observability + cost
#Security
- PII redaction · Prompt injection
- Rate limiting · Mandatory sandbox
- HITL approvals · Confirmation-gated tool
#Evaluation
- Eval suite · Evals CI
- Deterministic replay · Replay different model
- Prompt snapshots · Prompt diff
- Prompt experiments · Time-travel debug
#Apps + bots
#Skills + specs
#Conventions
- Replace
KEYwith your actual API key (useprocess.env.X_API_KEYin real code). - TypeScript; remove types for plain JS.
- Each recipe declares its
npm installline at the top.
Explore nearby
- PeerCustom adapter
Wrap any LLM API as an AgentsKit adapter. Plug-and-play with the rest of the kit in 30 lines.
- PeerAdapter contract tests
Verify any adapter against the ADR 0001 invariants A1–A10 with the shared test harness.
- PeerMore provider adapters
Mistral, Cohere, Together, Groq, Fireworks, OpenRouter, Hugging Face, LM Studio, vLLM, llama.cpp — one line each.