Wity Terminal: AI Agents in Your Command Line

Every developer has a browser tab problem. Not the general one — the specific one where a tool you actually need during your workflow lives in a browser, which means you have to break your terminal session, switch context, interact with a GUI, and switch back. For tools you use occasionally, this is fine. For AI agents you want woven into your actual build process, it's a real friction point.

Wity Terminal is the CLI package for Wity's agent infrastructure. Install it once, and your AI agents live in the same environment as the rest of your toolchain.

Installation

Getting started is one command:

npm install -g @wityai/agents-cli

That's the global install. Once it's in, you have access to the full Wity agent runtime from any directory in your terminal. Authenticate with your Wity account and your agents — including any you've built in Agent Studio — are immediately available.

What You Can Run from the CLI

The basic invocation pattern is straightforward: wity run [agent-name] with whatever parameters the agent expects. What this means in practice depends on which agents you've built or imported from the Agent Studio template library.

Research agents are the most common starting point. wity run competitor-monitor --urls domain1.com,domain2.com --output report.md — the agent crawls those URLs, identifies relevant changes or content, and writes a structured summary to a markdown file you can pipe wherever you need it. Run this from a cron job and you have automated competitive monitoring with zero ongoing maintenance.

Content generation agents work the same way. wity run pr-description --diff $(git diff main) passes the current diff to an agent that reads the changes and writes a structured pull request description — problem, changes made, testing notes. The output goes to stdout, ready to paste or pipe directly into your git workflow.

Code review agents can take a file path or a diff and return structured feedback: logic issues, edge cases, style inconsistencies, missing error handling. This doesn't replace human code review — it completes a first pass so that human review can focus on the things that actually require judgment.

Webhook Integration

The CLI is useful for manual invocation, but the real power is integration into automated workflows. Every agent you build in Wity has a webhook endpoint. From the CLI, you can register and trigger these webhooks directly, which means agents become part of your infrastructure rather than a separate tool you remember to use.

Common integration patterns:

• CI/CD pipeline hooks: trigger a documentation generation agent on every successful build. The agent reads your codebase, checks for functions missing docstrings, and opens a structured report. No manual step, no forgetting.
• Git hooks: a pre-commit hook that runs a security review agent on staged changes. If the agent flags something, the commit is held and the developer sees the issue before it goes into the repo.
• Monitoring integration: pipe your error logs into a log analysis agent via webhook. The agent looks for anomaly patterns, clusters related errors, and sends a structured digest to Slack or writes it to a file — whatever your alerting pipeline expects.

Chaining with Other CLI Tools

Because Wity Terminal outputs to stdout by default, it composes naturally with the rest of your Unix toolkit. The output of a research agent can go directly into jq for JSON parsing, grep for filtering, or another script entirely. This is the key architectural advantage over browser-based AI tools: composability.

A practical example: a developer maintains a script that checks five competitor blogs each morning, extracts new articles, runs them through a Wity summarisation agent, and formats the results into a daily briefing that arrives in their email. The whole pipeline is shell — the Wity CLI is just one step in a chain of tools that each do one thing well.

Real Production Use Cases

Automated documentation: on every deploy to staging, a GitHub Action triggers wity run generate-docs --source ./src. The agent reads the source directory, identifies public APIs, and writes or updates the documentation files. Docs stay current without anyone having to remember to update them.

Anomaly detection in logs: a nightly cron job pipes the day's error logs into a Wity analysis agent. The agent clusters errors by type, identifies new patterns that didn't appear in the previous 7 days, and writes a priority-ranked summary. The on-call engineer starts their day with a structured view of what changed overnight, not a raw log dump.

PR description automation: a git alias runs the PR description agent before pushing. Developers stop writing boilerplate descriptions and start writing better commit messages — the description handles itself.

Building Agents for CLI Use

Agents built in Wity's visual Agent Studio at wity.ai/tools/agent-builder are automatically available in the CLI. Build the agent graphically — define the steps, the prompts, the input parameters, the output format — and immediately run it from the terminal. The visual builder and the CLI runtime share the same agent definitions, so you're not maintaining two versions of anything.

For developers who want to define agents programmatically, the CLI also supports agent definitions as JSON config files, which means they can live in your repo, be version-controlled, and be deployed the same way as any other configuration.

The browser is a fine place to build agents. The terminal is where they should run.