AI agent

AI that works with your workspace.

Poly uses the workspace context you permit to create tasks, draft documents, search across your work and draw on boards.
Request a demoSee it in chat
Models
Bring your own
Streaming
Real-time
Memory
Persistent
Hosting
EU · self-host
Poly
online · gpt · claude
budget 38%
Summarize the #design channel and open tasks for what's left.
Searched messages & docs
Created 3 tasks
Writing summary doc

The design team shipped the new hero animation to staging and Lighthouse is at 100. Three items remain open — I created a task for each and assigned them to their owners. I'm writing the full summary to a doc

Design — Weekly Summary
Doc · just created
Ask Poly to do anything…
Search everything
Full-text across channels, docs and tasks.
Create tasks
Open, assign and update cards from a sentence.
Write docs
Draft and edit docs in place.
Draw on boards
Diagrams and flows via the board API.
Not training data
BYO keys — your content never trains foundation models.
Remembers
Persistent memory across your workspace.
Per-language prompts
Editable system prompts per language.
Morning brief
Workload insights, surfaced daily.
01It takes action

It doesn't just chat — it works.

Real tools inside your workspace

Poly doesn't only answer — it reads and changes your workspace. It creates and assigns tasks, drafts and edits documents, runs full-text search across everything, and even draws on boards — all through channel-aware tools that respect your permissions.

  • Create, assign & update tasks
  • Draft & edit documents
  • Full-text search across channels & docs
  • Draw diagrams on boards (board API)
Poly· working — 3 tools
budget 41%
Turn the review notes into tasks and sketch the release flow on the board.
Search docs
Create task
create_task
{ "title": "Fix hero contrast", "assignee": "selin" }
Draw on board

Opened 3 tasks from the notes and assigned owners. Sketching the release flow now

Fix hero contrast
Task · assigned to Selin
Release flow
Board · drawing…
02It knows the context

Persistent memory, daily brief.

No re-explaining every time

Poly remembers what matters. It keeps a persistent, per-user memory across your workspace and analyzes your workload to surface what's important in a daily brief. A screen-aware mode that takes what you're looking at as context is available as an experimental preview.

  • Persistent workspace memory
  • Daily brief with workload insights
  • ~22 tools across tasks, docs, boards & search
  • Screen-aware assistance (experimental)
Poly
Memory · on
Daily brief · Wed 09:00

2 tasks due today, the pricing doc got 4 new comments, retro at 15:00.

Where did we land on the pricing rollout?

Rollout is set for March 3. Legal signed off in Friday's email; the only open item is the price-table update — Deniz owns it.

Sources
Pricing v3 — NotesDocUpdate price tableTaskRe: rollout datesEmail
From memory: you announce pricing changes in #general first.
03Your models, your rules

Bring your own model.

Governance built in

Connect Anthropic, OpenAI, Gemini, or any OpenAI-compatible endpoint — including self-hosted models via Ollama. Providers are controlled by a workspace admin, per-model daily token budgets cap spend, and editable prompts let you tune Poly per language. Run it in our EU-region cloud or on your own servers — your content is never used to train foundation models.

  • Anthropic · OpenAI · Gemini · Ollama & compatible
  • Per-model daily token budgets
  • Editable, per-language prompts
  • EU-region cloud or self-host — never training data
AI providers
Managed by workspace admins
Add provider
ClaudeanthropicDefault
Enabled
claude-sonnet-4-5sk-ant-••••7f2q
Daily tokens62%
Llama — self-hostopenai_compatible
Enabled
llama3.3:70bhttp://ollama.internal:11434/v1
Daily tokens18%
Geminigemini
Disabled
gemini-2.5-pro
Never training dataEU cloud or self-hostPrompts per language

Architecture

Server-side, streamed.

01
Real-time
SSE token streaming
02
Persistent
Memory
03
Budgeted
Per model / day
04
Self-host
EU · your servers

Request a demo

Meet Poly.

An AI that lives inside your workspace — one that knows your context and does the work for you. Request a demo and we'll walk your team through it.

Request a demoSee it in chat