Build reliable agents with your coding tools.
WeHarness turns your project requirements into an agent architecture, helps build it in the tools you already use or in its own cloud sandboxes, and checks the behaviour with project-specific tests.
Start here
Section titled “Start here”IntroductionWhat WeHarness does, and what is available now.
Getting accessYour invitation, sign-in, two-step sign-in and recovery codes.
QuickstartFrom your workspace to a pull request in about ten minutes.
Use the app
Section titled “Use the app”SessionsOne conversation per project: attachments, slash commands, versions and the draft.
Plans and approvalHow a brief becomes a plan, and who decides what.
GitHubInstall the app, choose repositories and branches, push pull requests.
Credits and usagePackages, what uses credits, caps and budgets.
Build and check your agent
Section titled “Build and check your agent”Set up tracingPython, Node.js, plain OpenTelemetry and framework recipes.
EvaluationsTest sets, cases, evaluators, calibration and the release gate.
MonitoringYour agent over time, and alerts.
LessonsCorrections kept as rules, each shown by a test.
Use WeHarness from your coding tool
Section titled “Use WeHarness from your coding tool”MCP overviewWhat your coding tool can do, and the rules it follows.
Claude CodeAdd the server and sign in.
Codex, Cursor, VS Code, DevinSetup for each coding tool.
Tools and prompts referenceEvery MCP tool with its parameters, generated from the server.
Reference and trust
Section titled “Reference and trust”Python SDKweharness-tracing on PyPI.
Node.js SDK@weharness/tracing on npm.
SecurityAccounts, workspace isolation, sandboxes and where data is kept.
TroubleshootingSign-in, GitHub, sessions, credits and traces.