najmuzzaman-mohammad/gawkbot

open source grok bot. gawk bots automate your menial work via AI models and build you microapps to manage the outcome, so that you have a false sense of control.

gawkbot – Open‑source “always‑on” AI bots you run locally

What it is – gawkbot is a self‑hosted platform that lets you describe a workflow in plain language (or by demo) and automatically builds a dedicated AI‑driven “micro‑app” (a bot) to execute it. The bot runs on your own machine (or a VM you control) and can interact with 1200 + third‑party services via integrations. Every external write (e.g., sending an email, creating a CRM record, committing code) is gated behind an explicit human approval step.

Why it matters – Most commercial AI‑assistant products (e.g., xAI’s Grok Bot) are cloud‑only, closed‑source, and charge per‑use. gawkbot flips that model: the code is public under a Sustainable Use License, runs locally with your own API keys, and is free aside from the cost of the underlying LLM you choose (Claude Code, Codex, Opencode, Ollama, etc.). This gives you full control over data, costs, and security while still getting the convenience of AI‑generated automation.

Core concepts

Component	What you get
App (screen)	A live web UI (default http://localhost:7891) that shows the bot’s “screen”, its data, and a chat‑style console.
Routines	Scheduled jobs (cron‑like) that run the bot on a repeatable cadence; each run is version‑controlled and logged.
Tools	Custom actions you teach the bot (e.g., Score a lead, Post to #ae‑handoffs).
Knowledge	Auto‑generated wiki‑style pages that cite the source of every claim the bot makes.
Data + Integrations	Typed tables per bot plus a shared catalogue of 1200+ connectors (Slack, Gmail, HubSpot, etc.).
Approval gate	Reads are free; any write operation pauses until you click Approve in the UI.

Getting started (quick‑start)

Install – either run the one‑liner or install globally:

npx gawkbot               # opens the UI and runs the binary from npm
# or
npm install -g gawkbot && gawkbot

Choose a model – the first‑run screen asks for a provider. By default it uses Claude Code; you can switch to Codex, Opencode, Ollama, Hermes, OpenClaw, or any OpenAI‑compatible endpoint via --provider or environment variables.
Onboard – the UI walks you through runtime verification, naming your office (workspace), and describing your first workflow in a single sentence.
Watch the build – a live activity feed streams the bot’s screen, routine, tools, and knowledge pages as they are generated.
Approve – when the bot attempts its first external write, a modal asks you to confirm.

Building from source (requires Go 1.25+ and Bun):

git clone https://github.com/najmuzzaman-mohammad/gawkbot.git
cd gawkbot
cd web && bun install && bun run build && cd ..
go build -o gawkbot ./cmd/wuphf
./gawkbot

Extensibility & deployment options

Multiple runtimes – set --provider ollama (or any of the supported names) to point the bots at a local model or a custom OpenAI‑compatible endpoint.
Local‑first vs. cloud‑hosted – actions can run through a local CLI (your keys never leave the machine) or via Composio, which handles OAuth for the 1200+ SaaS integrations.
Workspaces – isolate projects with gawkbot workspace create|list|switch.
Team sharing – invite a teammate over a private Tailscale/WireGuard mesh (gawkbot share) or via a Cloudflare tunnel (one‑use, 6‑digit passcode).
Telemetry control – product analytics and session recordings are opt‑in; they are disabled unless you provide a PostHog key.

License & community

License – Sustainable Use License (see LICENSE). The repo is fully public; you can fork, modify, and self‑host without paying the original authors.
Community – Discord community linked from the README, plus a Hacker News badge indicating recent popularity.

Who should try it?

Engineers or small teams that want AI‑generated workflow automation without sending their data to a third‑party cloud.
Developers who need a programmable “bot per workflow” pattern and want full visibility into costs, token usage, and execution logs.
Anyone looking for a free, extensible alternative to proprietary “Grok Bot”‑style services.

Bottom line

gawkbot delivers a local, open‑source, approval‑gated AI bot platform that can turn a natural‑language description of a task into a runnable micro‑app, complete with scheduling, data tables, and a citation‑backed knowledge base. If you value data sovereignty, cost transparency, and the ability to inspect or modify the automation code, gawkbot is a solid foundation to build on.