What June can do in a reply

Every tool June can use while it answers, when it reaches for each, and where one is not available: the web and the app, Slack, or a scheduled run.

In a reply, June can do more than write: it searches, runs code, reads your connected tools and starts threads. Each use of a tool shows as a step in the reply that you can open.

June picks the tools from what you ask. You never have to name one, but it helps when you want one used ("search the web for…", "run it in your sandbox").

A tool step in a reply: open it to see what went in and what came back.

Every tool

Every tool works on the web and in the app: in your own chats, projects, organization chats and calls. What June may do with an integration can differ in an organization (personal or organization).

Find out

Tool	What June does with it	In Slack
Date and time	Checks today's date and the time	Yes
Web search	Searches the web and cites what it found. Not when an API key pays	Yes
Read a document	Reads a PDF, an image, a Google Doc or Slides deck, or a file in its sandbox or your project	No
Search your threads	Searches your threads, from the web and from Slack, for up to 30 matches	No
Databases	Runs read-only queries on your databases while a skill that uses one is loaded	No
Skills	Finds the skill for the task and loads its instructions	Yes

Make and run

Tool	What June does with it	In Slack
Run a command	Runs code and command-line tools in June's sandbox	No
Image generation	Makes an image and adds it to the reply, when a ChatGPT plan pays	No
Send a file back	Attaches a file it made in its sandbox to the reply, up to 20 MB	No
Save to Google Drive	Saves 1 to 10 files it made into your Drive, up to 20 MB each, with Google Drive connected	No

Act in your tools

Tool	What June does with it	In Slack
Your integrations	Reads and acts in the tools you connected on the Integrations page	Yes
Connect an integration	Checks whether an integration is ready, and gives you the link to connect it if not	Yes
Reminders and schedules	Sets up, lists, changes, pauses, resumes, runs and removes reminders and schedules	Yes

Hand off

Tool	What June does with it	In Slack
Start a thread	Starts a thread of its own for part of the work, when you ask for one	No
Ask one of your agents	Hands a task to one of your agents and brings back its answer, once you have an active agent	No
Design an agent	Drafts a new agent from the work you described	No

Remember

Tool	What June does with it	In Slack
Remember and forget	Saves a fact as a memory, or forgets one	Yes
Update the Soul	Adds to, clarifies or removes something from the Soul of the chat	No
Where it learned something	Finds the chats where a fact it knows was said	Yes

Your integrations. June reaches every connected integration through one tool: it writes a short script that searches your connections for the right action and calls it. Your passwords and tokens never enter that script. A run stops after 60 seconds, and its result is cut at 64 KB. What June may read or change in each service is on that integration's page.

Which of my Notion pages mention the Meyer GmbH contract?

Start a thread. June starts a thread only when you, the project's instructions or an agent's skills ask for a separate thread or parallel work. A question that needs research or depth is answered in the chat you asked it in. When you ask, the new thread can run as one of your agents, or on another model of your ChatGPT or Claude plan; June may also pick a lighter model of that plan for a light task, such as a lookup. June starts at most 5 threads in one reply. A child thread, linked under the one it came from, nests at most 3 levels deep; a side thread stands alone and has no such limit. At most 10 threads it started can run at once for you, and 30 in one organization. A thread June just started cannot start another in its first reply, though it can ask one of your agents. In an organization or a project, the new thread stays there.

Start a separate thread for each of the three suppliers and compare their offers

Ask one of your agents. June offers this only once you have an active agent, and knows your agents by name on the web and in the app. The agent answers from its own instructions and skills, in up to 8 steps, and June brings its answer back into your thread.

Image generation draws on a ChatGPT plan's image allowance, which runs out faster than text. The plan can be yours or one shared with the organization.

In a scheduled run

A schedule runs with only the tools it was set up with, in groups such as web search, integrations, documents or the sandbox. A schedule set up with none has only the date and time, and one June sets up from chat has none unless you ask June to turn some on by name: integrations, web, documents or terminal (what each opens). In a scheduled run June also cannot:

set up or change schedules or reminders,
generate images, send a file back to the thread, or save to Google Drive,
use databases, search your threads, start threads, ask an agent or update the Soul.

A scheduled run stops after 4 minutes without activity, and after 13 minutes at most.

What June does not do

It does not browse websites like a person, with clicks and forms. It searches the web and reads what search returns, or fetches a page from its sandbox.
It asks your approval only for a step that changes something in a connected account (what the card shows); reading never asks. In the sandbox it cannot ask yet, so a command that would need it does not run.
It does not send mail from a chat. In a mailbox it writes drafts for you to send. Only a scheduled run of one of your agents can send mail, and only through a sending connection that agent was granted; the Integrations page has no way to add one.