Skip to main content

​
Understanding Zen Agents and the Library

What are Zen Agents? Why are they useful?

Zen Agents are Zencoder’s bespoke copilots: you spin them up on the Agents page, give each one a friendly name plus a CLI alias, and pour in the exact rituals you expect— coding patterns, review rules, release etiquette, reporting beats. Their instructions sit alongside a curated toolbelt, so an observability agent might lean on search and file edit while an SRE tuner ropes in custom MCP actions. Once created they live in the Custom Agents section of the selector, ready for teams to reuse as the canonical way to run that workflow.

Watch video

How to browse the Agents Library?

The Library lives on a tab inside that same Agents page and acts like an inspiration gallery. Click the tab to see curated agents you have not installed yet, skim their thumbnails, and open any card to study its instructions, tools, and sharing defaults. Because each listing is editable after installation, you can treat the Library like a set of starter kits and choose the ones that match gaps in your team’s workflows.

How to use an agent from the Library?

When a Library agent looks promising, click through to its detail view and hit Add. The agent is copied into your Custom Agents list as if you authored it, so you can retune the instructions, swap tools, and decide whether to keep it personal or share it org-wide. From there it behaves like any other agent: select it in the agent picker when starting a conversation, gather feedback from the runs, and keep iterating on its instructions so the entire org benefits from each tweak.

How to identify use cases for new custom agents?

Agents shine when the work is repetitive, high-stakes, or steeped in tribal knowledge. Look for areas where senior engineers keep repeating the same instructions—code review checklists, secure-by-default patterns, release rituals—and translate those expectations into agent prompts instead of reinventing them every run. The more specific the mandate, the better: a deployment hardening agent or a “backend observability sweeper” will always outperform a single generalist. If a workflow is already documented in runbooks or captured in hallway conversations, it is mature enough to crystallize into a custom agent so the expertise scales beyond one person.

Share card via email

​
Building Your First Custom Agent

How to build a custom agent?

Start from the Agents page (three-dot menu → Agents) and hit Create agent. Give it a descriptive name for the UI, set an alias for CLI use, and then author the instructions: spell out the task scope, preferred coding patterns, review cadence, and reporting quirks so the agent knows exactly how to behave. The form preloads the core tools every coding agent needs—search files, read context, edit files—so once instructions feel right you can save and the agent will appear under Custom Agents in the selector.

Watch video

How to add tools to custom agents?

Below the instructions lives the tool roster. The defaults stay in place unless you remove them, and you can expand the toolbox by clicking Add tool, which opens a picker listing every MCP or integration already installed. Need something new? Choose Manage tools from that picker, configure additional MCPs on the Tools page, and jump back to the agent editor—the fresh capabilities now appear in the selector. Pair each tool with guidance in the instructions (“Use the Terraform MCP before editing infra files”) so the agent knows when to call it.

Share card via email

​
Scaling Expertise Across Teams

How to share a custom agent with the organization?

Every agent has a sharing dropdown: Personal keeps it private, Org promotes it to everyone. Flip that toggle before saving (or anytime later), and the agent surfaces under Custom Agents for every teammate alongside the creator’s email so folks know whom to ping. The same roster is visible in the web dashboard, where you can review both personal and org-wide agents; you may edit anything you created yourself but shared agents remain locked to their authors so governance stays tight.

Watch video

How to share a custom agent with the library?

The Agents Library is open source at https://github.com/zencoderai/zenagents-library under the MIT License. Fork the repo, add your agent definition following the provided template, and submit a pull request describing what problem it solves. Zencoder reviews every submission for usefulness and security; once approved, the agent automatically appears in the IDE’s Library tab so any customer can install it with the same Add button. It is the easiest way to share a proven workflow with the broader developer community while letting us handle distribution.

Share card via email