Keystroke

About Keystroke

Keystroke is an open-source platform for building AI agents and workflows. It launched this week in open alpha and is backed by Y Combinator. The platform is designed to take an agent from a text description to a deployed, working system inside a shared workspace.

Review

Keystroke enters a crowded field of agent-building tools, but it takes a different angle. Instead of just generating code, it tries to handle the operational side of agents: integrations, credentials, memory, and team collaboration. The platform is free to try with $20 in credits, which lowers the barrier for evaluation.

Key Features

• Describe an agent in natural language and the built-in agent builder creates it, connects tools, runs tests, and deploys it to a workspace.
• Agents ship with memory, web search, code execution, and persistent workspaces built in.
• Access to over 1,000 integrations, any API, or any MCP server.
• Workflows can combine agents with deterministic steps, multi-agent orchestration, and human approval gates.
• Teams interact with agents through chat in Keystroke, Slack, or Microsoft Teams, and share context and credentials from one workspace.
• Everything built on the platform is ordinary TypeScript, so it can be kept in git, reviewed, and run elsewhere.

Pricing and Value

Keystroke is currently in open alpha and free to try with $20 in credits. The company makes money from a cloud offering that has a free plan with usage-based pricing. The source code is available under an ELv2 license. Pricing beyond the initial credits and free tier is not yet fully defined, as the product is still in alpha.

Pros

• Open source with an ELv2 license, so teams can inspect and self-host if needed.
• The generated agents are plain TypeScript, which means no proprietary lock-in for the core logic.
• Built-in memory system stores facts as markdown files and conversations in a searchable database, inspired by research from Letta.
• Agents can run on schedules or app events and create their own triggers without extra plumbing.

Cons

• The "brains" knowledge base feature for spinning up vector indexes is still in testing with early users, not generally available.
• Because it's in open alpha, expect rough edges and breaking changes as the API and workflows evolve.
• Keystroke is not well suited for teams that need enterprise-grade governance and audit controls right now, since those details aren't documented yet.

Keystroke makes the most sense for small teams or startups that want to prototype internal agents quickly without stitching together separate frameworks, databases, and automation tools. The TypeScript output and open-source license mean you're not locked in if you outgrow the platform. For teams already invested in n8n or Zapier, it's worth a look as a different approach, but the alpha status means you should test it against your real workflows before relying on it.

Open 'Keystroke' Website