Overview
Cleo is an AI product manager that lives inside Telegram and Slack rather than another web dashboard. It learns how a team communicates, then automates the connective tissue work that usually falls between roles — running standups, tracking follow-ups, capturing decisions with their trade-offs, and even reviewing pull requests before merge.
Where most AI project management tools often create another workflow surface. Cleo instead positions itself as chat-first, plugging into the threads where work is already discussed. The pitch is operational, not organizational: keep using Telegram or Slack, and let an agent surface what was decided, what was promised, and what's behind schedule without anyone running a status report.
The product is designed around transparency rather than autonomy-by-default. Every learned fact ships with a source citation, a confidence score, and a correction option, and the user controls how aggressively Cleo acts through five trust levels — from passive observer up to a higher-authority operator mode that can act only after the team explicitly authorizes that trust level.
Key Features
- Standup automation in chat — Runs daily standups inside Telegram or Slack threads, collects updates, asks targeted follow-ups when something stalls, and posts a clean summary the team can scan in 30 seconds. No separate standup tool, no Loom catch-ups.
- Decision capture with trade-offs — When a discussion crosses a decision point, Cleo records the outcome plus the alternatives that were considered and the reasoning. Months later that decision is searchable instead of buried in scrollback.
- PR code review — Reads every pull request, flags likely bugs and risky changes before merge, and posts the review back into the dev chat alongside the standup updates. Works with AI agent coding workflows already on the team.
- Total Recall memory — Remembers every message, decision, and promise across the team's chats; surfaces what someone committed to last Tuesday when it becomes relevant today. Event-driven memory decay keeps fast-moving teams from drowning in stale facts.
- Transparent confidence scoring — Cleo's actions are gated by confidence: >85% acts silently, 60–85% surfaces a proposed action before doing it, <60% queues for review. No black-box automation that quietly invents the wrong meeting.
- Five trust levels — From observer mode (read-only, learning the team) to operator mode, where Cleo can take higher-authority actions after the team grants permission. The team graduates the level once each prior level has proven reliable; nothing automatic that wasn't authorized.
How It Compares
- vs. plain Slack/Telegram bots (Standuply, Geekbot) — Those are scripted standup collectors. Cleo combines standup automation with longer-running memory and decision capture; it also advertises PR review as a separate core feature.
- vs. Glean / Cassidy / general AI chiefs-of-staff — Glean, Cassidy, and broader AI chief-of-staff tools cover different knowledge-search and workflow-assistant use cases. Cleo should be framed more narrowly as chat-first PM operations rather than categorically deeper — it actually runs standups and acts on tickets through trust levels.
- vs. Linear / Notion AI built-ins — Linear and Notion are adding AI inside their products; Cleo deliberately sits outside, so the team isn't migrating data or workflow. Better for teams already invested in chat as the primary surface.
- vs. meeting-notes tools (Otter, Fireflies, Granola) — Those capture verbal meetings. Cleo captures async chat decisions and PR-driven work, which is where remote-async teams actually make most of their calls.
The honest read: Cleo is best when the team's real working surface is chat and the AI Chief of Staff category's broader offerings feel too heavy. It's a worse fit for teams whose decisions happen in calls or live in a Notion-style document hierarchy.
Pricing & Plans
| Plan | Price | Actions/mo | Best For |
|---|---|---|---|
| Launch free tier | $0 | 3 active automations, unlimited memory, per Product Hunt launch material; not shown as a separate plan on the official pricing section | Small teams or solo founders evaluating Cleo in Telegram |
| Pro | $10/mo | 500 actions | Active product teams running daily standups, decisions, and PR review |
| Ultra | $20/mo | 2,000 actions | Heavier usage — multiple repos, multiple chat channels, frequent automation triggers |
Product Hunt lists a 1-month free launch offer. Actions cover automated runs such as standup posts, PR reviews, decision captures, and follow-ups — sizing the plan depends on how chatty the team is and how many channels Cleo monitors. Slack is supported alongside the Telegram-first launch surface.
Best For
- Founding teams running on Telegram or Slack who don't want a Linear/Asana migration
- Indie-hacker and small-startup PMs who need standup, decision capture, and PR review in one agent
- Distributed/async teams where chat is the primary working surface
- Engineering managers who want a code-review safety net without adopting a separate review platform
- Operators of small businesses ready to compare against the broader AI assistant category before committing
FAQ
Is Cleo free?
Yes, in Telegram. The free tier includes 3 active automations and unlimited memory — enough for a small team to evaluate. Paid plans (Pro $10/mo for 500 actions, Ultra $20/mo for 2,000 actions) unlock heavier usage. Slack is described in launch material, but plan-level Slack access should be verified before relying on it.
What's an "action"?
An action is a Cleo-initiated event: a standup post, a follow-up question, a PR review, a captured decision, or any automated touch. The Pro and Ultra tiers count these against monthly limits.
Does Cleo work in Slack or only Telegram?
Both. Cleo is Telegram-first and is described as working with Slack; verify Slack feature parity and plan-level access before rollout. Roadmap and integration parity are still maturing given the alpha-stage status.
What other integrations exist?
Per Product Hunt's launch material, Cleo connects to Linear, Notion, Gmail, and Calendar in addition to Telegram and Slack. Specific feature depth per integration should be verified before relying on any single workflow.
What are the five trust levels?
A graduated authority model from observer (read-only, learning the team) up to operator (can move tickets, reply on behalf of the user). The team chooses where to sit; nothing escalates without explicit approval.
How does the confidence scoring work?
Cleo categorizes its own actions by certainty: above 85% acts silently, 60–85% surfaces a proposed action for approval, below 60% queues the item for human review. The aim is preventing the "confident-but-wrong" failure mode common in LLM agents.
Can I correct Cleo when it learns something wrong?
Yes — every learned fact ships with a source, a confidence score, and a correction option. Corrections are described as feeding back into Cleo's learned rules, reducing the chance of the same mistake recurring.
Is the PR review feature replacing a human code reviewer?
No. It flags likely bugs and risky changes before merge, but it's positioned as a safety-net layer, not a substitute for engineering review. Pair it with a senior reviewer rather than treating it as the final gate.
