Practical overview
When it fits
Use Deuz SDK when an application needs a TypeScript agent framework that combines model access with orchestration, durable execution, context features and operational controls.
Who it is for
- TypeScript developers building model-powered agents
- Teams orchestrating multi-agent workflows
- Builders needing resumable runs, approvals or persistent swarm state
- Developers integrating MCP tools, memory or retrieval into agents
Documented uses
- Connect agents to supported language-model providers through a common streaming interface
- Coordinate agent swarms with dependencies, runtime spawning and shared state
- Pause and resume workflows around human approval or verification
- Persist swarm execution state with SQLite or Postgres
- Add memory, retrieval, compaction and MCP tool calling
Requirements & limits
The repository is public and displays an MIT license. The documented setup requires Node.js 22+ or an edge runtime with Web APIs, npm and an Anthropic API key for the first-agent example.
- Provider choice: documented integrations include Anthropic, OpenAI, Google Gemini, xAI Grok, Mistral and DeepSeek.
- Execution model: swarms support dependencies, runtime spawning, shared blackboards and rounds.
- Persistence: SQLite or Postgres are documented options for swarm state.
- Operational controls: leases, cancellation, recovery, budgets, schedules and traces are included in the documented feature set.
- Runtime constraints: without a lease provider, each swarm run should be driven from one process.
Check current requirements in the official documentation. Source-based guidance does not mean OpenViST has run or independently tested this project.
Set up a first Deuz SDK agent
Install the core package and configure a first Anthropic-backed agent that returns a result.
Before you start
- Node.js 22+ or an edge runtime with Web APIs
- npm
- An Anthropic API key exposed as ANTHROPIC_API_KEY
Steps
Install @deuz-sdk/core with npm.
Set the ANTHROPIC_API_KEY environment variable.
Import runAgent and createAnthropic from the documented package paths.
Create an Anthropic model, pass it to runAgent with a prompt and maxSteps, then handle a completed result.
Expected outcome
A TypeScript agent is configured to request and print a completed model result.
Guide limitations
- The documented example uses Anthropic and does not establish configuration details for the other listed providers.
- The setup example does not document deployment or production execution procedures.
Compare options for Agent orchestration
These projects document a common task and share a project role. This is a source-based comparison, not an execution benchmark or a guarantee of interchangeability.
| Project | Documented purpose | Platform | Access and requirements |
|---|---|---|---|
| Deuz SDK | Connect agents to supported language-model providers through a common streaming interface | CLI, Cloud, Linux, macOS | The repository is public and displays an MIT license. The documented setup requires Node.js 22+ or an edge runtime with Web APIs, npm and an Anthropic API key for the first-agent example. |
| Agenite | Define typed tools with JSON Schema validation and error handling. | CLI | |
| Veldra | Create an agent with a policy, tools, skills, knowledge base and reasoning method. | Self-hosted | The documented deployment is self-hosted and uses a sign-in gate after a first-run administrator account is created. |
Sources & review
Source record details
- Source availability
- Active
- Sources checked
- 3 October 2026
- Original website title
- Deuz-AI/Deuz-SDK · Official homepage title · retained in its source language