Skip to content

Venture studio · AI agent platform

Conduity AI

AI agents for teams. They connect to tools such as Gmail, GitHub and Slack, reply in the channels the team already uses, and remember context between conversations.

Status
In development
Category
AI agent platform
Model
Internal venture

What it is

Many AI tools are a chat box. They cannot check a calendar, file an issue, send an email or remember what you told them yesterday, so teams connect several tools together and still do most of the work themselves.

Conduity agents connect to those tools directly, can run on a schedule, and keep context from one conversation to the next.

Who it is for

Product and engineering teams working in GitHub and Linear, operations teams that live in Slack and email, and anyone who wants an assistant that can take actions as well as answer questions.

Conduity is in development. If that sounds like your team, join the waitlist or book a demo.

What it does

Conduity is in development. This is what it is built to do.
  • Agents that take action

    Search email, create GitHub issues, schedule meetings, update Linear tickets, query Notion databases. Each agent has its own personality, toolset and knowledge base.
  • Every channel, one agent

    Deploy to web chat, Slack, WhatsApp, Telegram, email and SMS. The same agent, memory and context wherever the conversation happens.
  • Memory across conversations

    Agents remember facts, preferences, decisions and context across every interaction. The more you use them, the less you have to explain.
  • Scheduled and triggered automations

    Schedule daily standups, weekly reports or triggered workflows. GitHub and Linear webhooks can start an agent the moment something changes in the codebase or project board.
  • Multi-agent handoff

    Agents hand tasks to other, specialised agents. A general assistant routes a code question to the dev agent and a design question to the design agent, with context carried across.
  • Per-agent knowledge base

    Upload documents, sync Notion pages or ingest URLs. Assign knowledge to one agent or share it with all of them. Vector search finds the relevant information when an agent needs it.
  • Morning briefing

    A daily summary of calendar, open tasks, unread email and GitHub activity, delivered to your inbox or dashboard before the first meeting.
  • Agent-created automations

    Tell an agent to check on something every hour and it sets up the automation itself, without a configuration screen.

Integrations

Built-in connections, so there is no middleware or Zapier chain to set up and maintain.
  • Google

    Gmail search and send, Calendar events, availability checks, meeting prep.
  • GitHub

    Issues, pull requests, comments, webhook-triggered automations.
  • Linear

    Issue management, project tracking, event-driven workflows.
  • Notion

    Page search, database queries, content creation and updates.
  • Figma

    File metadata, components, comments, project browsing.
  • Slack

    Full bot integration with channel and DM support.
  • WhatsApp

    Business API messaging.
  • Telegram

    Bot integration with automatic webhook setup.
  • Email

    Inbound processing with auto-generated addresses.
  • SMS

    Messaging through Twilio.

How it works

  1. 01

    Create an agent

    Name it, describe its role, choose a model and assign the tools it needs. A system prompt sets how it behaves and writes.
  2. 02

    Connect your stack

    Sign in to Google, GitHub, Linear, Notion, Figma or whatever else the team uses, through OAuth.
  3. 03

    Deploy to channels

    Route the agent to Slack, email, WhatsApp, Telegram or your own web chat.
  4. 04

    Automate

    Set up scheduled briefings, cron-based workflows and event-triggered actions that run without anyone starting them.

Under the hood

  • Serverless

    Scales to zero and up to thousands, with no infrastructure to manage.
  • Encrypted credentials

    All credentials encrypted at rest with AES-256-GCM. Workspace-level data isolation.
  • Real-time streaming

    SSE streaming for immediate agent responses on every channel.
  • Vector memory

    pgvector embeddings for semantic recall and knowledge retrieval.
  • Background jobs

    An Inngest queue for automations, document processing and scheduled tasks.
  • Multi-tenant

    Workspace isolation with role-based access control, built for teams from the start.

Why we built it

Conduity is an internal venture, built, owned and operated by Outside The Box. We built it because chatbots could not do what teams need from them: connect to the tools they work in, remember context and run workflows.

Want early access?

hello@outsidethebox.dev
Conduity is in development; write to us or use the contact page to join the waitlist or book a demo.