It reads.
It writes.
It delivers.
Chatty AI builds a knowledge base from your content, connects live to the systems your teams run on, and finishes the job with a document you can send — all inside one conversation.
10
Knowledge
sources
14
Live
systems
Summarise last week’s incidents, open a Jira ticket for each one still unresolved, and give me the summary as a Word document.
Three incidents remain open. I raised a ticket for each and put the full write-up in incident-summary.docx.
01 · Knowledge base
What Chatty learns from
A knowledge base is the material Chatty reads: documents you upload, folders it keeps in sync, and repositories it connects to. Everything indexed here can be questioned in conversation and used to ground a model.
10
Files
Upload documents directly — the fastest way to give Chatty something to work from.
Folders
Upload an entire directory in one go, preserving its structure.
Archives
Drop in a zip or tar and Chatty unpacks and indexes what is inside.
Web pages
Point Chatty at a URL to pull a page into the knowledge base.
Text content
Paste text straight in when there is no file to upload.
SharePoint
Connect document libraries and sites, with source permissions carried onto every indexed fragment.
SMB network shares
Keep internal shares synced so the index matches what is on disk.
SFTP
Scheduled pickup from secure file transfer endpoints, indexed and ready to query.
FTP & FTPS
Read from legacy file servers that still drive day-to-day operations.
Directory sync
Watch a directory on a schedule so new and changed files are picked up on their own.
02 · In the chat
Systems Chatty acts in
A conversation reaches your whole knowledge base — and beyond it, into the systems your teams run on. These connections are live and two-way: Chatty reads current state and writes changes back, always as the person asking, never as a shared admin account.
Monday.com
Read boards and items, then create or update rows, owners and status from plain language.
Salesforce
Look up accounts, opportunities and cases, then create records, update fields and log activity without leaving the conversation.
Jira
Search issues, boards and sprints, open tickets, move them across the board and post comments.
Outlook
Read and search mail, draft replies in context and send on your behalf.
Calendar
Check availability across the organization, then schedule, move or cancel meetings.
Microsoft Teams
Read channel and chat context, then post messages, summaries and follow-ups back to the right thread.
SharePoint
Beyond indexing — write files back, update list items and manage document metadata.
OneDrive
Reach personal and shared files, and save generated documents straight back.
Excel
Read workbooks and worksheets, and write values, rows and tables back into them.
Microsoft Planner
Pull plans and buckets, create tasks and update assignments and due dates.
Microsoft To Do
Turn a conversation into tracked tasks, and read back what is still open.
OneNote
Search notebooks and sections, and append notes generated from a conversation.
Power BI & Fabric
Discover workspaces and semantic models, run DAX, surface reports in conversation and write measures and rows back.
SAP S/4HANA Cloud
Read business objects over OData and write changes back — bring the result straight into the conversation and act on it.
S/4HANA Cloud. SAP ECC is not supported.
More connectors are on the way
The list above is what is live today. New applications are added continuously — if something you rely on is missing, it may already be in development.
Or we build one for your systems
Connectors are not limited to off-the-shelf software. We develop them for the internal and industry-specific systems you already run — the same typed, permission-aware pattern as every connector above.
Tell us what you need connectedOne question, every system
Connect once, then ask across all of them
You choose which applications to connect, and connect as many as you need. Authorisation happens once and stays with you from one conversation to the next.
From then on you do not have to say which system holds the answer. Chatty queries everything you have connected and returns one picture — not six tabs you have to reconcile yourself.
Give me a status picture for today — my meetings against whatever is waiting on me in the work systems.
Checked 6 systems
Today
2 need attention
Two things need you before 11:00 — the sprint review has three open issues, and the cutover board is blocked on a dependency. Want me to chase both owners in Teams?
Automations
One trigger, several systems
Because Chatty writes as well as reads, a single event can carry all the way through — no copying between tools.
A customer escalation lands in Outlook
The ticket exists, the right people know, and nobody retyped anything.
A deal moves stage in SAP
Delivery sees it the moment finance does, with the numbers already attached.
Monday morning, before the leadership call
The pack is written and filed before anyone opens a spreadsheet.
A policy document changes in SharePoint
Everyone affected is told, and the review is already on the calendar.
And it ends with a file
Documents are written in the chat
The same conversation that searched your knowledge base and updated your systems finishes by producing a real Office file — ready to send, sign or file as it is.
Word
Reports, summaries and drafts as formatted .docx documents.
Excel
Tables, extracts and analyses as structured .xlsx workbooks.
Print-ready output for anything that leaves the organization.
Everyone sees only what they already have access to
Chatty acts as the person asking, never as a shared administrator. Every request carries that user’s own credentials, so a person can only ever reach what they were already permitted to open — in the source system and in the knowledge base alike. Nothing widens their access, and revoking it at the source revokes it here immediately.
Every integration uses standard OAuth 2.0 with delegated permissions. Your administrator consents once per application, each person then authorises individually, and tokens are encrypted at rest and refreshed automatically.
Your identity, not ours
Each person signs in once with their own Microsoft, Atlassian or monday account. Every call Chatty makes carries that user’s own token — there is no shared administrator account.
The source system decides
Because the request is made as the user, the vendor applies that user’s own permissions. If someone cannot open a SharePoint site in a browser, Chatty cannot open it either.
A fixed set of operations
Every integration exposes a defined list of operations over the Model Context Protocol. The assistant invokes what a connector publishes — it cannot construct arbitrary API calls.
Knowledge base permissions travel with the content. Source permissions are captured when a document is synced and applied at query time, so retrieval never returns a fragment from a document that person cannot read.
Write actions can require approval. Chatty includes a per-application confirmation step, enforced in the backend rather than by the model, that shows the exact payload for review and editing before anything is written. It is switched on per application as part of deployment.