Theo Agents

Draw the work.
Let Theo run it.

Build an AI agent the way you sketch a process: trigger, think, connect, wait for approval, and ship the finished artifact. Then watch the whole roster from one mission control.

Start creating freeSee the designer

One agent on Free. Unlimited agents on Pro and Teams.

Theo Agents

Everything Theo runs for you in the background. Schedules, webhooks, and on-demand flows. Quietly working, even when you aren't.

TemplatesNew Theo Agent
At a glanceLast synced just now
Total
3
Active
2/ 3
Next run
in 16h
Needs attention
0
Weekly Deep ResearchAutonomous

research · 4 steps

Track the market every Monday and land a sourced brief in Notes before the week starts.

Every Monday at 8:00 AMin 2d 14h12×
Daily Sales PulseAutonomous

sales · 3 steps

Read yesterday's pipeline movement and post the shortlist worth a call today.

Weekdays at 7:30 AMin 16h34×
Meeting Recap SenderWorkflow

operations · 4 steps

When a meeting ends, send every attendee the decisions and their own action items.

On meeting.completedNever run0×Ready to test
Lead routing✳Daily briefs✳Watchtowers✳Human gates✳CRM hygiene✳Content calendars✳Research reports✳Approval queues✳Org libraries✳Lead routing✳Daily briefs✳Watchtowers✳Human gates✳CRM hygiene✳Content calendars✳Research reports✳Approval queues✳Org libraries✳
Designer

The agent is a canvas, not a settings page.

Open one agent and the whole workspace becomes that agent: the graph, the setup panel, the run history, and Theo's builder rail. No global dashboard chrome fighting for attention.

Theo Agents DesignerTest runRuns
Theo

Agent builder

Ask Theo to change the graph

Theo converts plain English into the same edit operations the Designer uses. The graph stays visible and reviewable.

Mission Control

Every agent has a heartbeat.

Filter the roster by health or by trigger, sort by most failures, and let Theo raise its hand when something needs you. Three consecutive failures auto-pauses an agent, so a broken lane never keeps running quietly in the background.

Search agents…Recently active
1 Theo Agent needs attention. Auto-paused after 3 consecutive failures. Open them to inspect.
All6Active3Paused3Needs attention1Scheduled4On demand0Webhook0Event1Watchtowers1
Prospect OutreachAutonomous

sales · 4 steps

Work the new-lead list every morning and send the first touch in the prospect's own language.

Weekdays at 6:00 AMLast run 2d ago18×Auto-paused
Social Content CalendarWorkflow

marketing · 6 steps

Turn the week's shipped work into a scheduled board of posts, ready to review.

Every Friday at 9:00 AMLast run 6d ago26×
Competitor Page WatchtowerWorkflow

research · 0 steps

Notify-only: no steps at all, so a tripped monitor just rings the bell with what changed.

WatchtowerEvery 6 hours9×
Watchtowers

Some agents wait quietly until something changes.

A Watchtower checks a page, a search result, or a spreadsheet condition on a cadence. The first pass baselines. After that, Theo runs only when the signal actually moves.

No noisy polling

Quiet checks do not run the lane.

Changes are explicit

Every alert carries what moved.

Notify-only is valid

A Watchtower can be just a bell.

Pricing page changed

changed

competitor.com/pricing

Alerted the team and drafted a comparison note.

New Stripe payment

fired

payment.succeeded

Created onboarding tasks and sent the first email.

Sheet threshold crossed

waiting

Pipeline!B12 > 100k

Queued a review because approval is required.

Telegram approval

Approve before sending the deck?

Theo is about to email a 12-slide welcome deck to a new paid customer. Score: 87. Requested plan: Pro.

Human gates

Theo pauses when the decision should be yours.

Add an approval step anywhere in the lane. When the run reaches it, the agent waits, sends a compact summary, and continues only when someone says yes. That is how you ship AI work safely.

PresentationNotes docBoard cardEmailCode appResearch brief
Connectors

Connected services become steps.

OAuth once, then use the connection in every agent. Native connectors cover the common stack, and Portals can reach the websites with no public API.

Verified library

Start from an agent that already works.

The Library ships with 46 verified agents across two bundles — one that exercises OpenCharts itself, and one built around the departments you actually staff. Install one and you get your own fork: rename it, rewire the steps, change the schedule. The original never moves.

Publish your own back to the org and a partner can verify it. Verified agents can be routed straight from Theo Chat and called as a step inside another agent in the same org.

Verified badgeInstall as a forkCallable as a step

Library

Verified Theo Agents you can install or fork.

Search agents…

Browse agents

All46OpenCharts9Sales7Support6Marketing3Engineering3Operations2HR2IT1
Weekly Deep Research

OpenCharts

Run a multi-source deep-research pass on your topic, synthesize the findings into a structured brief with citations, and post it to the configured destination.

Autonomous128×
Install
Account Snapshot

Sales

Pull everything known about an account — open deals, recent tickets, last touch — into one page before the call.

Autonomous94×
Install
Ticket Triage

Support

Read every new ticket, set severity and product area, and route it to the queue that owns it.

Autonomous76×
Install
Release Notes

Engineering

Turn the merged pull requests since the last tag into customer-readable release notes.

Workflow61×
Install
Onboarding Plan

HR

Draft a 30-60-90 plan for a new hire from their role, their team, and the docs that already exist.

Autonomous43×
Install
Help Desk Triage

IT

Classify inbound IT requests, answer the ones with a known runbook, and escalate the rest with context attached.

Autonomous29×
Install

See it run

The canvas lights up as the agent works.

Run history, node status, credits, outputs, and approval waits are all visible on the same screen. You do not inspect logs after the fact. You watch the work happen.

5

trigger types

100%

visible runs

1

portable JSON

Theo Agents mission control with a live execution canvasTheo Agents
Questions

Agent questions, answered.

What is a Theo Agent?
A Theo Agent is a persistent AI worker that lives inside OpenCharts. You sketch the flow as a chain of steps, tell Theo what you want each step to do, and Theo runs the agent on a trigger of your choice. Every node maps to a real capability such as analysing data, scoring a lead, sending an email, or generating a presentation, and Theo executes them in order while passing data between steps in real time.
How is this different from Google Vertex AI Agent Builder, Zapier, or n8n?
Most agent and automation tools force you to write JSON, drag steps in a sidebar, or learn a custom DSL. Theo Agents is the only one where you literally draw the agent like a flowchart and click Run with Theo. It is AI-native (every node can use Theo), it ships with human approval gates by default, and it produces real deliverables such as presentations, podcasts, videos, research reports, and code apps. Flows export as portable JSON so they aren't locked in.
What can a Theo Agent do?
A Theo Agent can be triggered manually, on a schedule, by a webhook, or by a Telegram message. Once running it can call Theo for AI decisions or summaries, hit external services through connectors (Stripe, HubSpot, Slack, Google Sheets, Discord, Airtable, Calendly, Apollo, Typeform, Monday, and more), pause for human approval over Telegram, and produce deliverables: emails, slide decks, notes, podcasts, videos, code projects, content calendars, flashcards, and research documents.
Can an agent watch something and only run when it changes?
Yes. A Watchtower is an agent with a monitor trigger. It checks a web page, a web search, or a spreadsheet condition on a cadence, and only executes the rest of the flow when the signal actually changes. The first check records a baseline so you never get a false alert on setup, and a Watchtower can be notify-only if all you want is the alert itself.
Do I need to code?
No. You design the agent visually in the OpenCharts Designer, label each step in plain English, and click Run with Theo. Theo reads the labels, infers the right skill for each step, and auto-configures the flow. You can override any step manually if you want full control, or you can ask the Theo Builder rail in natural language and watch it update the canvas.
How does Theo decide which step to pick?
When you click Set up my flow, Theo reads your node labels, the surrounding context, and the data shape of the previous step. It then picks the best capability from the canonical step library (action, theo_action, path, approval, delay, call_agent). You can always override the inferred choice from the inline configuration form. Theo also surfaces its reasoning per step so you can see why a particular capability was selected.
How do human approval gates work?
Drop a Wait for Approval node anywhere in the flow. When execution reaches it, the node turns blue and you receive a Telegram notification with the data summary and Accept or Reject buttons. If you approve, the flow continues. If you reject, the run stops cleanly. Every approval is recorded in the run log along with the user who tapped Accept.
Can I run the same flow on a schedule?
Yes. Triggers include Manual (click Run), Schedule (cron-style: every Monday at 8am, every 2 hours, etc.), Webhook (an external HTTP request), Event (an in-platform signal such as a Stripe payment or a new task), Telegram (say "Run my lead flow" to the bot), and Monitor (a Watchtower that only fires when a page, search, or spreadsheet condition changes). You can change a flow's trigger from the Setup panel without rebuilding the graph.
Can I share Theo Agents with my team?
Yes. Set an agent's visibility to Org from the Designer header and it appears in your team's Library. Teammates can install a private fork with one click and re-installs are idempotent (you always get back the same fork instead of duplicates). An admin can also flip the Verified badge on a shared agent, which unlocks chat-side dispatch ("@daily-standup-brief, run now") and cross-agent calls from other Theo Agents in the same org.
What is the Library and how does verification work?
The Library is the org-wide catalogue of shared and verified Theo Agents. Owners and org admins flip the Verified badge from the Designer header. Verification unlocks two things at once: org-wide Library install and cross-user chat dispatch. Until an admin verifies an agent, it can only be dispatched by its owner. Cross-org dispatch is always refused.
Are Theo Agents free?
Every plan includes Theo Agents. The Free plan includes 1 agent, Starter includes 5, and Pro and Teams include unlimited agents. You get AI credits each month to run them; heavier or higher-frequency runs use credits from your plan, and Theo Agents share the same credit balance as everything else Theo does. There is no separate agent subscription.
Can I export and re-use my Theo Agents?
Yes. Every Theo Agent can be exported as a portable JSON file containing the graph, skill bindings, trigger config, and execution metadata. The format is open, so any compatible runtime can run these files. Import them later, share them with a teammate, or version-control them in Git.
How do I see what a Theo Agent is doing?
While a flow runs, every node lights up with its current state: yellow while running, green when done, red if failed, blue if waiting for approval. The Execution Log shows a timeline of every step with duration, credits used, and Theo's reasoning. You can click any node after the run to inspect its inputs and outputs, and the Insights tab aggregates run health, average latency, and feedback across the last 100 runs.
Soft clouds drifting across a bright sky

Build the agent you keep wishing existed

Start with a trigger, add the steps, then let Theo run the lane in public. Free to start and easy to share with your team.

Start creating freeRead the guide
Free plan includes one agent · Pro and Teams include unlimited agents