Practical overview
When it fits
Use Mozaik when an application needs interoperable agents with shared state, participant awareness, event-driven communication or concurrent agent loops.
Who it is for
- TypeScript developers building multi-agent applications
- Teams coordinating agents across projects
- Builders of concurrent, event-driven agent systems
- Developers needing shared agent context and runtime state
Documented uses
- Build agents that communicate across projects
- Run concurrent agents without blocking one another
- Coordinate humans, agents and observers in a shared runtime
- Use events and situation handlers to respond to changing runtime conditions
- Maintain shared typed state and agent memory context
Requirements & limits
The repository is public and identifies the project as MIT-licensed. Installation is documented through npm, Yarn and pnpm. Model-provider credentials are configured through environment variables.
- Mozaik is a framework runtime rather than a standalone end-user AI application.
- Its coordination model combines participants, situation handlers, shared typed state and published events.
- Agents can include instructions, tools and memory context, while observers can react without starting loops or sending messages.
- Provider credentials and model selection are part of the application setup.
- The runtime is written in TypeScript and runs on the server.
Check current requirements in the official documentation. Source-based guidance does not mean OpenViST has run or independently tested this project.
Set up a Mozaik agent runtime
Install the core package, configure a model provider and initialize the shared runtime state.
Before you start
- A TypeScript server project
- npm, Yarn or pnpm
- A provider credential for the model used by runInference
- The @mozaik-ai/core package
Steps
Install @mozaik-ai/core with npm, Yarn or pnpm.
Set the environment variable required by the provider you use, such as OPENAI_API_KEY, ANTHROPIC_API_KEY or GEMINI_API_KEY.
For DeepSeek models, set an OpenAI-style credential and base URL pointed at DeepSeek.
Define a runtime with defineRuntime, create a RuntimeState instance or subclass, and call initializeRuntime once with that state.
Expected outcome
The application has an initialized Mozaik runtime whose shared state can be resolved and used by participants and agent loops.
Guide limitations
- The documented setup does not by itself describe a completed agent workflow or a successful execution test.
- Calling initializeRuntime more than once throws, and using the other returned runtime functions before initialization throws.
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 |
|---|---|---|---|
| Mozaik | Build agents that communicate across projects | Server | The repository is public and identifies the project as MIT-licensed. Installation is documented through npm, Yarn and pnpm. Model-provider credentials are configured through environment variables. |
| 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
- jigjoy-ai/mozaik · Official homepage title · retained in its source language