Skip to content

Updates

Changelog

Release notes and version history for Smart Agents.

v1.0.1
April 2026release

On-premises API key auth and chat setup polish

First post-GA maintenance drop, shipped as AL build the current version. The focus is on-premises parity: on-prem BC tenants can now authenticate to the backend with a static API key instead of the secure authentication client-credential flow required in the cloud, and the setup page stops showing duplicate backend URL fields when the on-prem deployment mode is selected. The chat page also handles whitespace and blank API keys defensively so a mis-pasted credential fails fast at save time rather than on the first chat turn.

Added

  • On-premises deployment mode on Smart Agent Setup with static API key auth header against the QUALIA backend.
  • Defensive trim and empty-string validation on API key save, surfaced as inline errors on .

Changed

  • Smart Agent Chat page hides the duplicate backend URL field when on-prem deployment mode is active.
  • the backend service clears stale auth headers between requests so a rotated key takes effect immediately.

Fixed

  • 403 responses from the backend are now treated as non-retryable in CallApi, matching 401 behaviour and avoiding wasted credit reservations.
v1.0.0
April 2026release

Smart Agents 1.0 — General availability

Smart Agents for Business Central is now generally available on Microsoft AppSource, published by QUALIA Technik GmbH as AL extension build the current version against BC platform 26.0.0. Every new tenant is provisioned 500 credits on install, one-time, and can run agents across four channels — the in-product Smart Agent Chat page, Microsoft Teams, web chat, and email — from a single credit ledger and one audit trail. The four-tier model surface (Silver, Gold, Platinum, Frontier) is on by default, and every tool call is logged by tenant and user.

Added

  • AppSource listing under publisher QUALIA Technik GmbH, extension id 776a67fc-e35b-45f7-8e97-ec0bc6cb9077.
  • 32 templated agents provisioned on install via the templates module, covering finance, sales, inventory, HR, and back-office roles.
  • Four active channels: BC, Microsoft Teams, web chat, and email (the email handler polling the email platform).
  • Four-tier credit model (Silver, Gold, Platinum, Frontier) priced by the system credit system — 1 credit per 10k input tokens.
  • Full audit trail persisted in the tool call log and mirrored to Application Insights via the telemetry module.

Changed

  • Billing runs through Stripe prepaid top-ups and AppSource metered billing against the same credit ledger, replacing seat-based pricing.
  • Every new tenant receives 500 one-time starter credits on first install, seeded by the installer.
v0.9.0
March 2026update

Private beta for Microsoft CSP partners

Private beta drop for the first Microsoft CSP partner cohort, timed with the email channel going live and the move to the production backend. Partners can now scope agents per-tool, lock sessions to the caller's BC permissions, and read the full audit export for compliance sign-off. The four model tiers — Fast, Smart, Expert, Premium — were finalised in this drop and the dynamic tier pull from the backend was wired end-to-end, so a tier added on the backend shows up in BC without a new extension release.

Added

  • Email channel live in production via the email handler and Job Queue runner the email processor.
  • Data gateway the BC data gateway enforcing user-scoped OData reads against BC so agents never exceed the caller's permissions.
  • Audit export for compliance review sourced from the tool call log via the system (PDF and Excel).

Changed

  • BC extension now pulls model tier definitions dynamically from the QUALIA backend instead of hard-coding Fast/Smart/Expert/Premium.
  • Credit reservation moved to post-execution reconciliation against POST /api/v1/credits/reserve so actual token counts settle the hold.

Fixed

  • Token-refresh race in the backend service when a BC session expired mid tool call, resolved by treating 401 as non-retryable.
v0.5.0
March 2026update

First pilot tenants live

Early pilot milestone. The first tenants went live against their production BC APIs, running the core agent loop end-to-end against real ledger data. A smaller starter subset of the 32-template library covered the back-office work the pilots brought to the table — customer statements, aged receivables, and journal postings — so the chat-submit-and-poll loop was exercised on real finance data from day one. The backend ran SQLite with Litestream WAL replication to Azure Blob Storage, and the credit reserve endpoint was the single choke point for every chat turn.

Added

  • Core agent loop: the chat engine submits messages and polls the backend, with credit holds placed through POST /api/v1/credits/reserve.
  • Starter templates for customer statement generation, aged receivables reporting, and journal postings, pre-seeded in SmartAgent.
  • Backend persistence on SQLite with Litestream continuous WAL replication to Azure Blob Storage, per the Backend-Reference §Persistence.

Changed

  • Multi-step reporting jobs switched from a single-turn prompt to delegated sub-agent execution, collecting results from several agents in one session.