Back to directory

Deuz SDK

Deuz SDK is a zero-dependency TypeScript framework for building production AI agents, providing durable execution, long-term memory, hybrid RAG, MCP tool calling, human-in-the-loop approval and one streaming API across multiple model providers.

Deuz SDK

Deuz SDK is a TypeScript framework for building models, agents and swarms of agents in one SDK. Its core has no required runtime dependencies, with optional peer packages for specific integrations, and runs on Node.js 22+ as well as edge runtimes that support Web APIs.

It provides multiple model providers behind one streaming API, including Claude, GPT, Gemini, Grok, Mistral and DeepSeek, with structured output and tool calls. Agents get typed results, verification, approval steps and resumable runs, while swarms add task dependencies, runtime-spawned agents, shared blackboards, rounds and persistence to SQLite or Postgres. Additional modules cover operational concerns such as leases, draining, cross-process cancellation and recovery, persistent budgets and schedules, an evolutionary program-search mode with a mandatory budget, and memory, retrieval, compaction and MCP tools for context management.

It installs with a package manager and includes an installable coding-agent skill with API references and implementation guidance.

More in Agent frameworks

See all 902
  • Recurse

    recurse.run

    A serverless harness that lets a coding agent design, test, and deploy its own specialist sub-agents as tools, MCP servers, or bots, defined by a schema-validated manifest.

  • ProofForge

    shulgin.is-a.dev

    An AI-agent pipeline that decomposes a mathematics problem, proves each piece, and formalizes the result in Lean 4 and Mathlib until the Lean kernel accepts it, with six of its proofs already merged into Google DeepMind's formal-conjectures repository.

  • local-coder

    github.com

    Builds a local coding-agent team on top of Ollama and OpenCode by detecting the machine's hardware, recommending models for seven specialist roles, and testing whether each assigned model can actually use tools, edit a repository and pass verification.