> ## Documentation Index
> Fetch the complete documentation index at: https://help.casaconect.ai/llms.txt
> Use this file to discover all available pages before exploring further.
# Tools
> Everything the agent can do, grouped by area, with the risk tier that decides whether an action runs, is reviewed, or asks you first.
The agent acts through **tools**: small, well-defined operations such as *search the vault*, *edit a document* or *draft a reply*. It cannot do anything that is not a tool, and every tool enforces the same access rules as the rest of Casa Conect.
As the agent works, each tool call appears in the conversation as a short description, for example *"Searched your documents"* or *"Edited the document"*, so you can follow along.
## Risk tiers
Every tool is classified by what it can affect. The tier, together with your [permission level](/agent/permission-levels), decides what happens when the agent calls it.
| Tier | Meaning | Examples |
| :----------------- | :------------------------------------------------------------------------------------------------------------------------------------------ | :---------------------------------------------------------------- |
| **Read** | Looks, but changes nothing | Search, read a file, list matters |
| **Additive** | Creates something new, or makes a tracked edit you can undo | Create a document, edit with tracked changes, add a note |
| **Sensitive** | Changes or removes existing records, widens who can see them, reaches outside the organization, reconfigures the agent, or drives a browser | Delete an item, update a matter, reply to an email, write a skill |
| **Own checkpoint** | Has its own built-in approval step, whatever your level | Populate, submitting a plan, building a Shared Space |
A tool that has not been classified is treated as **sensitive**.
## The catalogue
| Tool | What it does | Tier |
| :-------------------------------- | :------------------------------------------------ | :------------- |
| Search the vault | Find passages across the documents you can access | Read |
| Read a file | Read a stored file | Read |
| List a vault | Browse folders and files | Read |
| Web search | Look up public facts | Read |
| Ask the user | Put one to five questions to you | Own checkpoint |
| List projects and members | See projects and who is in them | Read |
| List inbox documents | See documents emailed to your docs address | Read |
| List and read email conversations | See Inbox conversations | Read |
| Tool | What it does | Tier |
| :--------------- | :----------------------------------- | :------------- |
| Create a project | Start a new project | Additive |
| Update a project | Change a project's details | Sensitive |
| Enter a project | Move the conversation into a project | Own checkpoint |
| Tool | What it does | Tier |
| :--------------------------- | :------------------------------------ | :-------- |
| Create a folder | Add a folder | Additive |
| Rename, move or copy an item | Reorganize existing items | Sensitive |
| File an inbox document | Move an emailed document into a vault | Sensitive |
| Delete an item | Permanently delete a file or folder | Sensitive |
The agent is instructed to confirm before deleting anything it did not create in the same conversation.
| Tool | What it does | Tier |
| :---------------------------------- | :---------------------------------------------------------------------------------- | :------------- |
| Create a document | New editor document | Additive |
| Create a Word file | New `.docx` | Additive |
| Open a Word file | Open a `.docx` for editing. Changes become native tracked changes in a new version. | Read |
| Read a document, outline or section | Read long documents efficiently | Read |
| Search within a document | Find passages in one document | Read |
| Write or edit a document | Edits to existing content arrive as [tracked changes](/documents/tracked-changes) | Additive |
| Add a comment | Comment on a passage | Additive |
| Populate a document | Run the [Populate](/workflows/populate) workflow | Own checkpoint |
| Tool | What it does | Tier |
| :------------------------------------------ | :----------------------------------------- | :-------- |
| List and read matters, labels and templates | Look things up | Read |
| Create a matter | Open a new matter | Additive |
| Add a note | Write to the matter's activity | Additive |
| Link a reference or a conversation | Connect documents and email to the matter | Additive |
| Update a matter | Change status, priority, dates or assignee | Sensitive |
| Set labels, maintain lists | Reorganize matter data | Sensitive |
| Apply a template | Restructure the description | Sensitive |
Matter tools exist only when Matters is enabled for your organization.
| Tool | What it does | Tier |
| :--------------------- | :--------------------------------------------------------- | :------------- |
| List a thread's emails | Read the conversation | Read |
| File email attachments | Save attachments to a vault | Additive |
| Reply to an email | Draft, or send under your [reply policy](/inbox/approvals) | Sensitive |
| Forward an email | Always drafted for approval | Sensitive |
| Propose a change | Suggest an update to a record, for approval | Own checkpoint |
Replies and forwards to anyone who is not a member of your organization always become drafts for a member to approve.
| Tool | What it does | Tier |
| :--------------- | :----------------------------------------------------------- | :-------- |
| Send to a thread | Brief another conversation or a matter, or report back to it | Sensitive |
A hand-off arrives in the other conversation as its own turn, with the message shown as data rather than as instructions. Chains are limited to three hand-offs.
| Tool | What it does | Tier |
| :------------------------------------------------- | :--------------------------------- | :------------- |
| List and read spaces, templates and document slots | Look things up | Read |
| Create a Shared Space | Set up a new space | Sensitive |
| Build a Shared Space | Populate a space from a template | Own checkpoint |
| File to a Shared Space | Put a document into the right slot | Own checkpoint |
Anything that gives an outside party access shows you an approval card first, at every permission level.
| Tool | What it does | Tier |
| :---------------------------------- | :---------------------------------------------- | :------------- |
| Open a session, navigate, act | Drive a real browser | Sensitive |
| Observe, extract, take a screenshot | Read the page | Read |
| Request takeover | Hand the browser to you, for example to sign in | Own checkpoint |
| Close the session | End the browser session | Own checkpoint |
See [Browser sessions](/agent/browser-sessions).
| Tool | What it does | Tier |
| :------------------- | :---------------------------------------- | :-------- |
| Find and read skills | Look up a playbook | Read |
| Write a skill | Create or update a [skill](/agent/skills) | Sensitive |
The agent cannot delete skills. Only people can, from Settings.
| Tool | What it does | Tier |
| :---------------------- | :--------------------------------- | :------------- |
| Keep a task list | Track the steps of a larger job | Own checkpoint |
| Write and submit a plan | Propose a plan for your approval | Own checkpoint |
| Start a helper | Run a read-only helper in parallel | Read |
## Tools by situation
The agent does not always have every tool. The set narrows as the situation becomes less supervised.
| Situation | Tools available |
| :------------------------------------------------------------------------ | :------------------------------------------------------------------------------------------------------------------------------------- |
| **You, chatting in the dashboard or Word** | Everything above |
| **Plan mode** | Read tools and planning only, until you approve the plan |
| **A helper** | Read tools only. No edits, no questions, no helpers of its own. |
| **A conversation that came from email, even when you are chatting in it** | No delete, no skill writing, no browser, no helpers, no Shared Space creation |
| **Email from a verified member, unattended** | Vault reading and organizing, Word editing, Populate, matter work. None of the exclusions above are lifted. |
| **Email from an allowed external sender** | Triage only: the thread's own emails, filing, linking, drafting. No vault search, no file contents, no member lists, no Shared Spaces. |
The exclusions for email are structural. Those tools are not given to the session at all, so no instruction, however cleverly phrased in a message, can invoke them. See [Agent safety](/security/agent-safety).