Practical overview
When it fits
Use Zypher Agent when building an application that needs a configurable, tool-using agent whose task events can be handled as they arrive.
Who it is for
- TypeScript developers embedding AI agents in applications
- Teams connecting agents to filesystem, search, terminal or custom tools
- Developers integrating MCP server capabilities or handling streamed agent progress
Documented uses
- Configure an agent with a supported Claude or OpenAI model and selected tools.
- Run a task and consume its progress as streamed events.
- Expose MCP server capabilities to an agent.
- Track, review and revert agent changes using Git-based checkpoints.
Requirements & limits
The documented installation path uses Deno and JSR. The README says npm support is coming soon. The repository is licensed under Apache-2.0.
- The documented provider families are Anthropic Claude and OpenAI GPT.
- The example installs through JSR using Deno and imports rxjs-for-await to consume events.
- Tool options include file operations, search, terminal commands and custom tools.
- The framework documents native MCP server support with OAuth authentication.
- Git-based checkpoints support tracking, reviewing and reverting agent changes.
Check current requirements in the official documentation. Source-based guidance does not mean OpenViST has run or independently tested this project.
Install Zypher Agent and configure a task-streaming example
Set up the documented Deno/JSR SDK example with filesystem tools and an MCP server, then consume task events.
Before you start
- Deno
- The Zypher Agent package from JSR
- The rxjs-for-await dependency used by the example
- A supported model provider and model configuration
Steps
Install the package with
deno add jsr:@zypher/agent.Import
createZypherAgent,createFileSystemToolsandeachValueFromas shown in the README.Configure the agent with a model, filesystem tools and an MCP server.
Call
agent.runTask(...)and iterate over the resulting events witheachValueFrom.
Expected outcome
The example configures an agent for a task and processes its events as they stream.
Guide limitations
- The README’s example uses a Claude model; OpenAI GPT is also named as supported, but other providers are not established.
- The example demonstrates event consumption, not a user-interface implementation.
- The README says npm support is coming soon.
Sources & review
Source record details
- Source availability
- Active
- Sources checked
- 16 October 2026
- Original website title
- corespeed-io/zypher-agent · Official homepage title · retained in its source language

