AI agents that work with your apps

Build AI workflows on a visual canvas

Wire agents to Slack, Telegram, GitHub, your database and a RAG knowledge base, start them from a chat message or a click, and watch every step run.

agentflow://slack-support-bot
TriggerAgentAppsRun
Trigger
Slack mention
@AgentFlow in #support
Agent
Support agent
GPT-4.1 Mini
App
GitHub
Look up open issues
Output
Reply in chat
Answers in the thread
Run
4 nodes completed in 2.8s

Works with the apps you already use

OpenAI
pgvector
Slack
Telegram
GitHub
PostgreSQL

Features

Everything you need to put agents to work.

Canvas-first workflows

Wire triggers, agents, databases, apps and outputs on a visual canvas, and choose which steps run together and in what order.

Agents with real tools

Agents run on OpenAI models and can search the web, read pages, call HTTP APIs query your PostgreSQL database read-only and create PDF reports.

RAG knowledge bases

Add documents to a knowledge base in your own Postgres (a free Neon database works). They are chunked and embedded with pgvector, and agents search them by meaning.

Start from a chat

Mention your bot in Slack or message it on Telegram to start a workflow, and get the answer back in the same conversation.

Run history

See what started each run, every step’s output and the tools agents called. Runs from chat light up on the canvas as they happen.

Your keys, encrypted

Bring your own OpenAI and Tavily keys. Keys and app tokens are encrypted at rest and only used to run your workflows.

How it Works

From idea to a working agent in one flow.

01

Build the workflow

Drag a trigger, agents and outputs onto the canvas, and wire in the apps and databases each agent may use.

$ trigger: slack_message
$ agent: support_agent
$ knowledge: support_docs
02

Connect your accounts

Add your OpenAI key connect a Postgres (Neon) database, Slack, Telegram or GitHub, then choose what each agent is allowed to do.

$ openai: connected
$ pgvector: neon
$ github: read + comment
03

Run and watch

Click Run or let a chat message start it, follow each step live on the canvas, and review the results in History.

$ run: started from slack
$ steps: 3 of 3 done
$ output: replied in thread

Product

Build, run and review in one place.

The canvas, run history and node settings sit side by side, so you can change a workflow and see what it did without switching screens.

Trigger
Telegram message
Agent
Research assistant
Knowledge
RAG docs
Output
Reply in chat
Database
Read-only query

Build your first agent workflow today.

Start with a blank canvas, connect your OpenAI key and the apps you use, and hand repeatable work to an agent.