The Tool Signal
Hands-on reviews of AI tools that actually help small businesses

Astra Now Controls Calendar, Slack, ClickUp by Voice?

Updated 16 September 2026 · ai tools, voice ai, automation, openai

Astra Now Controls Calendar, Slack, ClickUp by Voice?

The headline going around says Astra now controls your calendar, Slack, ClickUp and email by voice. That is about half right. You talk to Astra's voice mode, and it hands the instruction to Codex — OpenAI's separate tool that writes code and carries out multi-step computer tasks on its own. The calendar and Slack control that everyone is sharing online came from a custom dashboard, built specifically for a launch demo. There is no button you can press today to buy a finished, one-click version of it.

That distinction decides whether this is worth your afternoon. If you were hoping to pay $20 a month, say out loud "move my Thursday calls to Friday and tell the client on Slack," and watch it happen — the evidence available right now does not support that. That kind of one-command control needs someone to build it for you first: a developer setting up and maintaining a small custom project inside Codex, wired to your specific calendar, Slack and task accounts. If you have that person, or can hire one, the ceiling is genuinely high: the demo showed real events being written to a real Google Calendar, not a mockup.

What Astra voice mode actually is

Astra voice mode only listens and passes on what you say — Codex is the part that actually does the work. A detailed build writeup published by MindStudio in September 2026 describes it as a spoken interface sitting on top of Codex, OpenAI's separate agent for writing code and finishing multi-step tasks unattended. You describe a task out loud. Codex then starts a separate background process to carry it out — similar to handing a task to an assistant and letting them go work on it in another room. You can keep talking to Astra, or walk away, while that happens.

That setup matters more than the voice part. Voice is just how the instruction arrives. Whether anything actually reaches your calendar depends on what the underlying Codex project was built to connect to: which accounts it has been given the logins and permissions for, and what a developer wrote into it to tell it what to do. Nothing in the available sources describes those setup steps. Treat the connection work — linking each account and telling the system what it is allowed to touch — as something you, or a developer you hire, will have to figure out from scratch.

What the demo showed it doing with calendar, Slack, ClickUp and email by voice

The demo dashboard did four concrete things, and all four involved writing, not only reading. According to the MindStudio writeup and a separate social walkthrough of the same launch material, the dashboard pulled live calendar data, Slack, ClickUp, email and Fireflies meeting notes into one interface. From there the operator could:

  • Create real calendar events. Edits made in the dashboard reflected back to the user's actual Google Calendar, so the calendar was functional rather than a mock.
  • Open and answer Slack conversations. Messages could be read and replied to from inside the interface.
  • Read and update ClickUp tasks. New tasks could be created from the same place.
  • Handle email threads. Existing threads could be opened and answered, and new ones started.

Two limits sit right next to those wins. First, the demo only worked with Google Calendar, Slack, ClickUp and specific mail accounts — no source claims Astra voice mode works with any calendar, task app or mail provider in general, so Outlook, Teams and Asana users have nothing to go on yet. Second, the whole thing was a one-off project, built for one person's exact set of tools, described in the source material as an "Ops OS" build. It is not a feature you switch on. Somebody had to design and build it from scratch, the way a business might hire someone to build a custom website rather than buying one off the shelf.

What it costs right now

The subscription floor for Astra Pro inside the desktop tooling looks like roughly $100 per month per seat, as of early September 2026. Below that, access exists, but on tighter allowances.

Plan Monthly price What is reported about Astra access Reported by, and when
ChatGPT Plus $20 Astra access included in existing allowances, mainly through Work and Codex, with credits available to buy 9to5Mac, 4 September 2026
Pro (Codex tier) $100 Astra Pro included in Work and Codex; one summary estimates about 50 Astra chat messages a week, shared with Sol Pro igmguru and a Japanese corporate guide, 10 to 15 September 2026
Pro $200 Astra included with roughly 200 chat messages a week by one author's estimate, plus the largest single seat allowance in Work and Codex agency plan guide, 4 September 2026
Business price not given in our sources Astra rolled out to Business alongside Pro 9to5Mac, 4 September 2026
Enterprise price not given in our sources Astra Pro available once a workspace admin enables it 9to5Mac, 4 September 2026

Every figure in that table should be re-checked on the official pricing page before you commit money. The weekly message counts in particular are author estimates from third party guides, and they are presented that way in the source material. They are not OpenAI's published numbers.

The API rate card

There is a second way to pay for Astra, instead of a monthly plan: the API, a direct pay-as-you-go connection that developers wire into their own software rather than a chat app. If a business goes this route, the published rate card matters more than any subscription price.

Astra reads and writes in "tokens" — small chunks of text, roughly three-quarters of a word each. Guides from ofox.ai and uravation.com, both dated 15 to 16 September 2026, report Astra priced at $10 per 1 million input tokens (what it reads) and $50 per 1 million output tokens (what it writes back). Sending the same background information more than once gets expensive fast, so Astra offers a cheaper cached input rate of $1 per 1 million tokens for text it has already seen, with a $12.50 per 1 million charge to store new context in that cache. Push past roughly 272,000 tokens of input in one go, and pricing moves to a long-context band: $20 per 1 million input tokens and $75 per 1 million output tokens, with cached input at $2 and cache storage at $25 per 1 million.

A separate review comparing AI providers reports that OpenAI's own model-comparison page lists Astra at that same $10 input / $50 output rate, and its smaller sibling model, Sol, at $4 per 1 million input and $20 per 1 million output — an independent match against OpenAI's own listing rather than another aggregator repeating the same figure.

Output costs five times what input costs. That ratio is the single most useful number here. An assistant that listens to a long day of context and answers in one line is cheap. One that drafts twelve email replies is not.

Credits inside Work and Codex

Work and Codex bill in credits rather than dollars. An agency oriented guide from early September 2026 lists 250 credits per 1M input tokens, 25 credits per 1M cached input, and 1,250 credits per 1M output tokens, and notes that 25 credits comes out near $1, which lands the effective cost back at the same $10 and $50. The same guide says Fast mode costs 2.5 times the standard credit rate in Work and Codex, against 2 times in the API.

The limit that will actually bite

A launch tracker citing OpenAI's own pricing page estimates gives local Astra message capacity per rolling five hour window, current as of 3 to 16 September 2026: 5 to 45 messages on Plus and Business Standard, 25 to 225 on Pro at $100, and 100 to 900 on Pro at $200. The ranges are wide because message cost varies with context size.

Read the bottom of each range, not the top. On a $20 Plus seat, five voice instructions in five hours is a plausible bad afternoon. That is not a schedule you can run a workday around.

Three honest problems with this idea today

There is no product here, only materials. Every working example in the available sources is a custom dashboard someone built with Codex and a set of connected accounts. No source describes a turnkey, non technical personal OS for freelancers. If you cannot maintain a small project, or pay someone who can, you are buying a promise about what is possible.

The cost is hard to forecast. You are looking at a subscription, plus credits, plus a rolling allowance, plus a token price where output runs five times input. A quiet week and a busy week will not cost anything like the same. Compare that to the flat per seat pricing on most of the AI email assistants that already sit in a small team's inbox, where you know the bill before the month starts.

Write access to four systems is a real blast radius. A voice instruction that reaches your calendar, your Slack, your task board and your outgoing mail is an agent acting under your name in front of clients and colleagues. A misheard sentence does not stay private. Nothing in the sources describes the confirmation or undo behaviour, so assume you will have to design that part yourself.

Who should skip this for now

Skip it if your stack is not the demo stack. The documented integrations are Google Calendar, Slack, ClickUp and specific mail accounts. Running Outlook, Teams or Notion means you are building on top of nothing that has been shown to work.

Skip it if you are a solo operator with no build capacity. The gap between "Codex can do this" and "this runs my week" is a project, and it is a project that needs upkeep every time a permission expires.

Skip it if a predictable monthly number matters more than a high ceiling. At $20 a month the allowances are thin, and at $100 to $200 a month you are paying senior contractor money for a workflow you still have to assemble.

The cheaper path that works this week

Most of what people want from this is not really voice. It is the reply that gets drafted, the task that gets created, the meeting that gets moved. A conventional automation tool does those today, for a fixed price, with a visible log of what it did. Our comparison of n8n and Zapier for small business automation covers where each one lands on cost and maintenance.

The meeting notes piece is already solved too. The demo dashboard pulled Fireflies notes in as one of its data sources, and the same category of tool works perfectly well on its own, as covered in our breakdown of how the main AI meeting note takers compare.

Build the boring version first. If it turns out the only thing missing is saying it out loud instead of typing it, you will know exactly what Astra needs to add.

FAQ

Does Astra control your calendar, Slack, ClickUp and email by voice, as the headlines claim?

No source we found supports that. The demonstrated control came from a custom dashboard built with Codex and connected accounts, described in the September 2026 build writeups as an Ops OS project. Astra plus voice mode is the platform layer. The integration layer is something a person configures.

What is the cheapest plan that gets me Astra?

ChatGPT Plus at $20 a month, according to 9to5Mac on 4 September 2026, with Astra usage included in existing allowances and extra credits available to buy. The catch is capacity: the per five hour estimate for Plus runs as low as 5 Astra messages. Check the current pricing page before you sign up, since this rollout is moving weekly.

Will it work with Outlook, Teams or Asana?

Nothing in the current sources says either way. The demos are tied to Google Calendar, Slack, ClickUp and specific mail accounts, and no vendor claim of universal calendar or mail support appears in the material we reviewed. Treat non demo tools as unproven until someone shows a working build.

How much would a month of real use cost on the API?

It depends almost entirely on how much text Astra writes back. At the reported rate of $10 per 1M input and $50 per 1M output tokens as of 16 September 2026, reading a lot and answering briefly is cheap, and drafting long replies all day is not. Cached input at $1 per 1M makes repeated context far cheaper, so a well built project reuses context instead of resending it.

Is the $200 Pro plan worth it over the $100 one?

Only if you are hitting the ceiling on the cheaper tier. The reported difference is capacity rather than capability: roughly 25 to 225 local Astra messages per five hours at $100, against 100 to 900 at $200, with those figures dated 3 to 16 September 2026. Start lower and watch where you actually run out.

Verdict

The capability is real and the product is not. A dashboard that reads and writes a live Google Calendar, answers Slack, updates ClickUp tasks and sends email from a spoken instruction was demonstrated in September 2026, and that is a genuinely useful ceiling. Getting there means a Codex project, connected accounts, ongoing maintenance, and somewhere between $20 and $200 a month before token or credit costs. For a small business with no build capacity, the honest move is to wait for the packaged version and spend the next quarter automating the same jobs with tools that already have a price list.

Related reading

About the author

The Tool Signal is written by a small team that runs automation work for client businesses every day, mostly on Google Workspace, Slack and a task board of some kind, which is exactly the stack in these demos. Astra has not been in our hands, and this article says so in the claims block at the bottom of the page: every price and limit above is attributed to the publication that reported it and the date it was reported, so you can check it yourself. We will update this page when somebody publishes a real setup guide. Until then, treat any headline saying Astra now controls your calendar, Slack, ClickUp and email by voice as a demo you would have to rebuild yourself.

Latest updates

What we checked, and what is the vendor's word

  • Astra: Astra is listed at $10 per 1M input tokens and $50 per 1M output tokens in OpenAI's API catalog. Not verified by us, taken from a third-party page (16 September 2026, source).
  • Astra: Once input passes roughly 272,000 tokens, Astra moves to a long context band priced at $20 per 1M input and $75 per 1M output. Not verified by us, taken from a third-party page (16 September 2026, source).
  • Codex: Inside Work and Codex, Astra is billed in credits at 250 credits per 1M input tokens and 1,250 credits per 1M output tokens, with Fast mode costing 2.5 times the standard rate. Not verified by us, taken from a third-party page (16 September 2026, source).
  • ChatGPT: Astra rolled out first to Business and Pro customers on the $100 and $200 per month plans, with Plus users added later. Not verified by us, taken from a third-party page (16 September 2026, source).
  • Codex: Local Astra messages in Work and Codex are estimated at 5 to 45 per five hours on Plus, 25 to 225 on Pro $100, and 100 to 900 on Pro $200. Not verified by us, taken from a third-party page (16 September 2026, source).
  • ChatGPT: A Pro tier at $100 per month is listed as including Astra Pro access inside Work and Codex. Not verified by us, taken from a third-party page (16 September 2026, source).
  • Astra: Astra voice mode is described as a spoken interface layered on top of Codex, which opens a background thread to carry out the task. Not verified by us, taken from a third-party page (16 September 2026, source).
  • Astra: In the launch demo, a custom dashboard pulled in live calendar, Slack, ClickUp, email and Fireflies notes, and events created in it appeared in the user's real Google Calendar. Not verified by us, taken from a third-party page (16 September 2026, source).

Prices and limits move. Each line above says the date we last saw it on the source page.