aideas.org

Already delivered (6)

Advanced Excel processing
July 2026
Work with complex Excel files and export calculation results back to Excel.
Microsoft SSO
August 2026
Register and sign in with your company account.
Update to the privacy shield
August 2026
Personal data is tokenized before it reaches commercial model providers — you always get the originals back. Polish, Italian and English recognizers; the filter stops traffic rather than letting data through.
Unified model layer
August 2026
New models and providers can be added centrally, with per-project budgets and cost visibility.
Answer quality evaluation service
August 2026
Back-end foundation for systematic, repeatable quality measurement of agent outputs. This is the groundwork the rest of the evaluation work sits on — measuring live agents from the interface will follow as a separate feature.
Live reasoning & tool-call streaming
August 2026
Watch the agent’s reasoning, its tool calls and their results as they happen, instead of waiting for a finished answer.

Quarterly model refresh
1st week of September · recurring
Gemini 3.7 Flash and GPT 5.6 available in the app.
Microsoft SharePoint connection
September
Use SharePoint as a source of files in chat and as a source for your knowledge base.
Agent orchestration
September
A durable execution engine underneath the agents: long-running tasks survive infrastructure restarts and resume from the last checkpoint, with strict isolation between projects.
Elicitation — the agent asks before it acts
September
For complex or ambiguous tasks the agent asks two or three targeted follow-up questions instead of guessing what you meant.
Guided presentation generation
September
The agent reads your source material, agrees the outline with you in conversation, and only then builds the deck — and you can keep refining the same outline without starting over.
Security monitoring & tamper-evident evidence store
from September, rolling
Centralized security logging with automated alerting and immutable retention, documented for enterprise security reviews.

Agent orchestration, Elicitation and Guided presentation generation build on each other, in that order.

Long-running task orchestrator
October
Multi-step runs with pauses, resume after failure, per-run cost budgets, schedules and notifications.

Answer quality loop
November
The agent plans, executes and then reviews its own output against a quality threshold before returning it, with hard cost caps on every loop.

No committed date.

Kimi K3 hosted by Nebius

Open-weight models running on EU infrastructure, as an alternative processing location for projects that require it.
MCP Servers Connection

Connecting the app to external tools and data sources through the Model Context Protocol.
Knowledge graph & wider company context

A temporal knowledge layer on top of your documents, so the agent can work with relationships and company-wide context, not just individual files.
Charts, diagrams and interactive widgets in chat

Visual components rendered inline instead of walls of text; delivered incrementally.