Skip to main content
Keystroke gives you a code-first toolkit for building AI agents and workflow automations, and a managed platform to run them. Not sure where to start? Use the sections below to find the features that best match what you’re trying to do.

Build a specialized agent

An agent is an LLM you give a job, a model, and a set of tools. It reasons about a request, calls the tools you expose, and works toward a result. Reach for an agent when the steps aren’t fixed and you want the model to figure out the path. Define agents in TypeScript, give them tools built from your actions, choose a model, and add memory, subagents, or sandboxes as your use case grows. See Agents and Build agents.

Automate multi-step work with workflows

A workflow chains actions and agents into a durable, multi-step automation. Reach for a workflow when the steps are known and you want them to run reliably and predictably, the opposite end of the spectrum from letting an agent decide. Workflows orchestrate the deterministic parts of your automation and can call agents for the parts that need reasoning. See Workflows and Build workflows.

Connect the tools your team already uses

Keystroke ships with over 1,000 built-in integrations, so your agents and workflows can interact with the tools your team relies on. You can also connect to any HTTP API (including private internal APIs), wire up MCP servers, and manage OAuth and API keys in one place. See Integrations and Credentials.

Run automatically on a schedule or event

A trigger kicks off a workflow without anyone pressing a button. Run automations on a schedule with cron, in response to a webhook, or from app events like a Google Meet call ending or a new Linear issue. Triggers attach to workflows and filter on the source, so each automation runs exactly when it should. See Triggers.

Talk to your agents in Slack

Wire an agent to Slack and other chat surfaces so your team can interact with it where they already work. Ask questions, kick off tasks, and get results back in a thread, with no separate UI to learn. See External channels.

Build it all with your coding agent

Because everything is TypeScript in your repo, your coding agent (Cursor, Claude Code, Codex, etc) can build and maintain your project for you. It reads AGENTS.md and the bundled skills to learn how Keystroke works, then writes agents, workflows, and triggers on your behalf. See AI onboarding and llms.txt.

Deploy and operate on the platform

The Keystroke platform is a managed place for your team to run what you build. Deploy your project from TypeScript, then inspect run history, manage credentials, invite members, and set access controls, all from one shared workspace.

Learn more

Quickstart

Install the CLI, scaffold a project, and deploy your first agent.

Agents

Define LLM agents with tools, models, memory, and sandboxes.

Workflows

Chain actions and agents into durable, multi-step automations.

Integrations

Connect Slack, Google, Linear, and 1,000+ other tools.