onKaulonKaul is an open-source AI developer platform with a pluggable intelligence core — investigate incidents in Slack, triage Jira tickets, build in the browser with an embedded coding agent, and pair-program with teammates in real time.
Connect Sentry, Datadog, GitHub, Jira, and Confluence once. Then ask from anywhere — a Slack thread, a Jira comment, the CLI, or the web UI. The bee-worker queue handles the rest asynchronously, so your team never waits for a response.
onKaul meets you where you already work — no new tools required.
Mention @onkaul in any Slack thread. The bee-worker picks it up, runs a full investigation (Sentry → Datadog → code), and replies in the thread with citations.
Comment @onkaul on any Jira ticket. onKaul reads the issue, correlates logs and errors, and posts a formatted reply using Atlassian's ADF — proper headings, code blocks, bold.
Run uv run onkaul for a local interactive shell. No webhooks, no Slack — just you and the agent. Comes with a setup wizard to configure all integrations.
onkaul> prompt/setup wizard for all integrationsA full browser-based chat interface with streaming SSE responses, persistent session history, and multi-user support. Works out of the box with your repo config.
Every sandbox is an isolated Docker container. Open it, start the coding agent in the terminal, and describe what to build. The live preview updates on every file save. Share a link — your collaborator joins the same session in real time.
Opens with bypassPermissions — reads, writes, and runs files immediately. No confirmation prompts.
SSE file watcher triggers iframe reload on every save. Preview stays in sync with the agent's edits.
Share a token URL. Guest gets the same live preview + joins the same tmux session — real-time pair coding.
9 presets from 4K to Galaxy S24. Each viewer picks independently — no sync required.
Creates a new branch, commits all changes, pushes, and opens a GitHub PR — from a single button.
Static HTML, Vite/React, or Fullstack (Vite + FastAPI). Or connect an existing GitHub repo.
Drop images and files (up to 20 MB each) directly into the sandbox. The agent can reference them immediately.
Start from scratch, then link a GitHub repo. Or open an existing configured repo directly.
Each sandbox runs in its own Docker container. No host access — the agent operates inside the container only.
When you @mention onKaul in Slack or Jira, the request is queued via Redis RQ and picked up by a bee-worker process. Workers run investigations asynchronously — your thread gets a response when it's ready, without blocking anything.
Scale horizontally by running more workers. Each worker processes one job at a time and logs everything to JSONL for observability.
Pick a configured repo or create a new project (Static, Vite, or Fullstack). Container starts in seconds.
Start the coding agent in the terminal. Describe what you want. Preview updates live on every save.
Send a share link for live pair-coding, or push to a new GitHub branch and open a PR — right from the UI.
Run /setup in the CLI or set env vars. Add Slack, Jira, Sentry, Datadog, GitHub, Confluence.
Mention @onkaul in Slack or Jira. A bee worker picks it up, runs the investigation, and replies.
Approve a plan, then let the agent write the code, push a branch, and open a PR automatically.