Autobot

Autobot is AutoSage's built-in AI assistant. It chats with you, generates scripts and workflows, helps you reason about your automation library, and reads from your existing setup to give grounded answers — all from inside your AutoSage account.

Think of it as a copilot for automation. You describe what you want in plain language; Autobot writes the script or drafts the workflow, saves it into your library with one click, and explains its reasoning along the way. With your own LLM key connected, it can go further — running your automation, watching it stream live, and helping you diagnose and fix failures.

This section is for using Autobot

These pages walk through Autobot from a user's perspective. If you're after the architecture (service topology, internal APIs, deploy pipeline), that lives in the product repo's engineering docs, not here.

What Autobot can do

• Chat about automation — ask how to approach a problem, get an explanation of a script you already have, or brainstorm a workflow design.
• Generate scripts — Python, PowerShell, or shell. Autobot writes them and saves them straight into your Script library, ready to drop into a workflow.
• Generate workflows — describe what you want ("every weekday at 7am, run the cleanup script on staging-1, then email me"), and Autobot builds the node/edge graph and saves it as a draft.
• Run scripts and workflows — in execution mode (with your own LLM key), Autobot can launch a run, watch it stream live in the Run Panel, investigate failures, and propose a fix.
• Read your existing library — it can list your scripts, workflows, vaults, servers, and credentials so its suggestions match what you already have. It references vault resources by name and id only — it never sees secret values.
• Stream responses in real time — answers appear token-by-token, and tool calls (like "creating script…") show up inline as they happen.
• Remember the conversation — each thread keeps full history. Older turns get summarized automatically when a chat gets long, so it stays coherent without blowing the model's context window.

Where Autobot draws the line

Some boundaries are deliberate, not missing features:

• Execution is gated. Autobot can run automation, but only in execution mode, and only when you've connected your own LLM key. Reading and generating are always available; running is something you switch into on purpose.
• It will not reveal vault secrets to you or to the model. Credentials are referenced by id only, and a runtime password travels browser-to-server directly — never through Autobot.
• It does not manage cloud infrastructure, write Ansible playbooks, or upload files. (Cloud-infra tooling is on the roadmap.)

Autobot acts under your identity

Autobot can only touch resources you could reach yourself through the normal UI, and it can't escalate its own privileges. Running automation — the most consequential action — sits behind an explicit mode switch and your own LLM key. See Limits & Privacy for the full reasoning.

---

Getting started

Here's how to go from "where is it?" to "it just built something for me" in three steps.

1. Open Autobot

From the left navigation, click Autobot. You'll land on the chat surface.

• Start a new chat — type into the input at the bottom. Your first message creates a new thread, and the thread title is auto-generated from your prompt.
• Open an existing chat — pick one from the History panel on the left.

There are two Autobot entries in the left navigation: Autobot (the chat surface) and Autobot Dashboard (your usage analytics).

2. Ask for what you want

Autobot understands plain language. The more specific you are about your environment, the better the result. Some prompts that work well:

"Write me a Python script that pings a list of servers and prints any that are unreachable."

"Show me the workflows I have and tell me which ones don't have a trigger configured."

"Build a workflow: every weekday at 7am UTC, run db-backup.sh on prod-db-1, and email me when it finishes."

"What does the script cleanup-logs.py actually do? Walk me through it."

"I have a credential called aws-readonly — which workflows use it?"

When Autobot performs an action — creating a script, building a workflow, listing your library — you'll see an inline badge (e.g. "Creating script: ping_servers.py") and the result appears immediately afterward. The mechanics of those badges and the live token stream are covered in Chat interface.

Mention names of things that exist

Script names, workflow names, credential names — say them. Autobot can read your library, but only when prompted to. "Run on prod-db-1" works far better than "run on the database server", because Autobot can look prod-db-1 up in your Vault and wire in the right credential.

3. Find what Autobot built

Whatever Autobot creates lands in the same place you'd find it if you built it by hand — it isn't tucked away in the chat.

• Scripts land in your normal Script library. Open the Script Editor and they'll be there, owned by you. See Script actions.
• Workflows land on your Workflows page as drafts. Open them in the workflow builder to review the node graph before running them. See Workflow actions.

Trust, but verify

Autobot generates first drafts. Review a script body before running it, and open a generated workflow in the builder to check the node graph and triggers. The chat is a copilot, not autopilot. Even when you let Autobot run something, you stay in the loop — runs are opt-in, metered, and you confirm before they fire.

Tips for good results

• Be specific about your environment. Name the exact server, the exact credential, the exact constraint ("UTC only", "Python over shell").
• Iterate. Ask for a draft, then refine in the same thread — "now add error handling", "use the aws-readonly credential instead". Each turn builds on the last.
• For running and long debugging sessions, use BYO. Execution mode requires your own provider key, and tool-using chats (especially workflow generation) can burn tokens quickly. The shared admin quota is a starter; heavy users should plug in their own key.

How this section is organized

Page	What it covers
Chat interface	The chat surface, streaming, tool badges, modes, threads, history, and archive.
Script actions	How Autobot reads, writes, saves, and runs scripts.
Workflow actions	How Autobot drafts, saves, and runs workflow graphs.
Execution copilot	Running, watching, investigating, and fixing runs — modes, the Run Panel, and password safety.
Customizing & LLM keys	Tone, expertise, language, custom instructions, per-thread overrides, and connecting your own provider key (BYO).
The Autobot Dashboard	Usage buckets, model breakdown, and your remaining quota.
Limits & Privacy	Quotas, rate limits, privacy, security, and troubleshooting.