ChloeGet Started

Compare

Eight ways to build an AI agent, side by side.

The eight things Chloe was built around, and which of the others have them. Every cell about another project comes from its own docs.

Compare Chloe with
Chloe and seven other ways to build an AI agent, marked on eight things, with what each is best for.
  • Has it
  • Partly
  • Not there
ChloeOpenClawby the OpenClaw FoundationMastraby Mastraeveby VercelGoogle ADKby GoogleLangChain.jsby LangChainOpenAI Agents SDKby OpenAIClaude Agent SDKby Anthropic
Get startedSide by sideSide by sideSide by sideTheir docsSide by sideTheir docsTheir docs
At a glanceA dot for each row below, filled when it has it.
8 of 8
6 of 8
7 of 8
3 of 8
3 of 8
4 of 8
1 of 8
3 of 8
Written in codeThe agent is a program you can read, test and keep in git.Yes. TypeScriptNo. Markdown files and settingsYes. TypeScriptPartly. Markdown and TypeScriptYes. TypeScript, Python and moreYes. TypeScript or PythonYes. TypeScript or PythonYes. TypeScript or Python
Mix code, AI and agentsPlain code, one typed answer from an AI, and an agent with tools, in one job, in the order your code says.Yes. Built in, per stepPartly. Through add-onsYes. In workflowsPartly. Only in tools the model callsYes. With workflowsYes. As steps in a graphPartly. In your own code around itPartly. Your code, between agent runs
Any modelAny AI provider with one key, or a subscription you already pay for, and a model per job.Yes. Any, by key or subscriptionYes. Many, local ones tooYes. Any, from 200+ providersYes. Any, through AI GatewayPartly. Gemini first in TypeScriptYes. Any providerPartly. OpenAI first, others in betaNo. Claude only
Runs on a scheduleEvery morning at 8, or every 15 minutes, with nothing else to set up.Yes. Built in, in your time zoneYes. Built inYes. Built in, with a databaseYes. Built inNo. Use Google Cloud SchedulerPartly. Only on paid LangSmithNo. Not built inNo. Not in the SDK
Shows the priceWhat each run cost you, in dollars, with no other service to sign up for.Yes. Per step, and code steps are freeYes. Per replyYes. Per runPartly. Only through Vercel's AI GatewayPartly. Counts usage, shows no pricePartly. Only in LangSmithPartly. Counts usage, shows no priceYes. Per run, estimated
Comes with a dashboardA page to watch every run: what it did, and what it cost.Yes. On your machine, or onlineYes. Local web app, freeYes. Studio, local or hostedPartly. On Vercel, paidPartly. For testing, or Google CloudPartly. LangSmith, hostedPartly. Traces, on OpenAI's siteNo. Not built in
Memory you can readWhat it remembers is Markdown files you can open, change, or read in a dashboard.Yes. Markdown in git, and in the dashboardYes. Markdown files you can editPartly. Markdown, in a databasePartly. Markdown, in Vercel BlobNo. Google Cloud or a databasePartly. Database, or Deep Agents filesPartly. Chat history, sandbox notesYes. Markdown files on your disk
Yours to keepOpen source, on your own machine, and the runs and memory stay there.Yes. MIT, all on your machineYes. MIT, on your machineYes. Apache 2.0, some parts paidYes. Apache 2.0, on your own serverYes. Apache 2.0, Gemini firstYes. MIT, no account neededPartly. MIT, traces go to OpenAIPartly. Closed source, Claude only
Best forDaily routines, in codeA personal assistant in chatFull-stack agent appsChat agents on VercelAgents on Google CloudComplex graph agentsMulti-agent apps on OpenAIClaude Code inside your app

Read from each project's own docs on 7 Oct 2026: OpenClaw, Mastra, eve, Google ADK, LangChain.js, OpenAI Agents SDK and Claude Agent SDK. Out of date? Tell us.

One at a time

Chloe beside one of them, row by row.

Chloe
$ npm install @chloejs/core && npx chloe setup