Release Notes

A detailed record of every release, including new features, improvements, and bug fixes. The newest release is listed first.

---

​

v0.17.0 — Folder organization, remote NAS storage, and a redesigned Settings experience

​

Added

• Upload and organize files into folders — across Projects, Agents, Chat, and Workflow Builder — with drag-and-drop and tree navigation.
• Connect a NAS storage location on a remote host over SSH, with password, private-key, or IP-allowlist authentication.
• Agents now generate better-designed Word documents and PowerPoint decks by default, with a refreshed visual style.

​

Improved

• Settings, Admin, and list pages like Agents and Projects now share a cleaner, consistent header and tab design.
• The Backup and Restore admin page gained search, sorting, and filtering for policies and snapshots, plus a clearer restore-progress view.
• Agents in large workspaces respond faster, correctly drop files or databases you’ve deselected, and can name a source file’s folder.
• System emails like password resets and invites now show your logo and consistent brand colors.

​

Fixed

• Adding an agent you own back into a project no longer fails with a permissions error.
• Editing an agent nested inside a project no longer bumps it out of the project — breadcrumbs and back-navigation stay put.
• Long agent conversations no longer show a generic “not available” error — a busy workspace now shows a clear retry message instead.
• Disabling a feature like Skills or a connector for your team now also blocks the AI from using it.
• Emails sent by agents through Gmail or Outlook now render numbered lists and tables correctly instead of raw formatting characters.
• Opening a file in Google Docs, Slides, or Sheets now reconnects automatically after an expired connection, or prompts you when Drive access is missing.
• Storage location cards no longer overlap or misalign on narrower screens.

---

​

v0.16.0 — Self-serve billing & credits, SCIM provisioning, and enterprise audit trails

​

Added

• Self-serve billing — subscribe to a plan with a card, manage seats, and cancel anytime, right from Settings.
• Admins can buy extra usage credits and assign them to individual team members, with a dashboard showing who’s using what.
• SCIM provisioning connects your identity provider to automatically create, update, and deactivate accounts and sync teams.
• A new audit log tracks account activity — logins, role changes, invitations — with export to your SIEM tool (webhook, Splunk, or Datadog).
• A post-login terms-of-service prompt now appears for users who haven’t yet accepted the latest agreement.
• API and cURL command snippets now copy correctly for your shell — Bash, PowerShell, or Command Prompt.
• Usage history can now be exported as a CSV with date-range and activity-type filters.

​

Improved

• Billing & Usage settings were redesigned with clearer plan comparisons, a credits dashboard, and easier seat management.
• Usage charts now show a rolling 30-day view instead of resetting to zero on the 1st of each month.
• The Activity log was reworked with clearer columns and a full CSV export with filters.
• Deactivated users are now hidden from the Users list by default, shown only with a clear status badge.
• SuperAgent’s side panel can now be resized by dragging, and the message box stops growing after 10 lines.

​

Fixed

• Fixed several billing issues — seat limits now match your plan after upgrading, adding seats no longer errors, and assigned credits show up for members.
• Web-search citations in chat now render as compact numbered reference chips instead of raw text links.
• On touch devices, pressing Enter in chat now adds a new line instead of sending your message early — tap Send to send.
• The Artifacts panel no longer overlaps the chat header, and project Agent cards no longer overlap each other.
• Custom skill uploads are more reliable, including macOS zip exports and skills that reference restricted commands.
• Document processing is more reliable: scanned PDFs are now searchable, storage-scan citations are accurate, and stuck indexing jobs after project deletion are fixed.

---

​

v0.15.0 — Activity logs, real-time notifications, and change password

​

Added

• Activity log and notification center added — track your actions and agent runs, receive real-time alerts for quota limits and sharing, and get desktop notifications even while the app is open.
• Change your password from Settings → Account — the form validates your new password against the policy rules live before you save.

​

Improved

• Stopping a streaming response now freezes content instantly, and tool blocks no longer replay from the start when you collapse and re-expand them.
• Navigating after a new deployment no longer shows “Something went wrong” — the app now recovers automatically from stale-bundle errors.

​

Fixed

• Generated dashboards now render radial charts correctly, titles display as expected, and deep-link scroll inside generated views is restored.
• Express-mode files no longer show citation markers in chat — they were incorrectly appearing after the agent finished responding.
• Settings page headers now stay pinned while you scroll, so the section title stays visible as you move through a long settings page.

---

​

v0.14.0 — Single Sign-On, generative dashboards, and faster everything

​

Added

• Single Sign-On is here — connect your identity provider and log in without a password.
• SAML 2.0 support added alongside OIDC, so corporate identity providers can be wired up.
• SSO group-to-team mapping: your identity provider’s groups automatically translate into Generate teams on first login.
• One login, many workspaces — switch between tenants from the profile menu without signing in again.
• Agents now generate interactive dashboards, charts, forms, and visualizations inline in the chat panel, not just text.
• Generated dashboards and forms can be downloaded as standalone HTML files that work offline.
• Search your chat history by message content, not just conversation title.
• Scheduled agent runs now receive the job’s title, description, and attached data sources, so they act on the right context automatically.

​

Improved

• Chat responses use a cleaner visual hierarchy — headings, lists, and dividers now match a modern document style.
• Agent answers are structured answer-first: the key point comes before supporting detail, with less unnecessary scaffolding.
• The Add Files panel now shows segmented upload-mode toggles and per-file icons so you can see exactly how each file will be processed.
• The Admin Users and Teams tables now support sorting and filtering so large workspaces are easier to manage.
• Core API response times are significantly faster — reduced redundant work on the most-hit endpoints.
• Pressing Enter in a chat edit now sends the updated message; Shift+Enter inserts a newline as expected.
• The SSO configuration wizard guides admins through setup step-by-step, with live connection testing before saving.

​

Fixed

• Chat history no longer loses its scroll position or flickers when navigating long conversations.
• Agents no longer crash mid-run when a tool returns an unexpected argument format.
• Inviting a user who is already a member now shows a clear duplicate-email error instead of failing silently.
• The browser tool used by agents now recovers from state-mismatch errors instead of getting stuck.
• The chat stream no longer leaves a blank spacer at the top of the panel after a response finishes.

---

​

v0.11.1 — Reliability polish for chat and PDF extraction

​

Fixed

• PDFs with unreadable PUA text layers now force OCR so extraction no longer fails silently.
• Agent service compaction is now triggered based on the model’s context window instead of a hardcoded threshold.

---

​

v0.11.0 — Chat resilience, Cerner FHIR, and richer agent history

​

Added

• Cerner FHIR R4 connector with automatic OAuth token refresh — query EHR data alongside your other sources.
• Chat now survives network drops: SSE streams resume from where they left off when the connection comes back.
• Agent conversations now keep a full history checkpoint so you can resume long sessions without losing context.
• New browser tool features for agents that drive web pages on your behalf.

​

Fixed

• Sandboxed agents no longer refuse new requests while a cold-create is in flight — they wait for it instead.
• Editing the first turn of a chat now resumes the correct conversation thread.
• Workflow generator now uses per-caller authentication, so generated workflows respect the requesting user’s permissions.

---

​

v0.10.0 — Custom MCP servers, Phase 1 Security, and a more reliable email stack

​

Added

• Bring-your-own MCP servers: add custom MCP servers from the Connectors page, grant them to specific agents, and use their tools at runtime.
• Upload custom skills as a ZIP file — no more copy-pasting individual files.
• Trigger agents on a schedule or from external HTTP calls.
• Notifications when an MCP server connects, disconnects, or surfaces an error.
• Skill subcategory display in the catalog, plus friendlier validation when a skill fails to load.
• Phase 1 security hardening: stronger authentication, token hashing at rest, TLS preparation, and automatic PII scrubbing in logs.
• Centralized email service so password resets and invites don’t get throttled by Gmail’s IP-based blocks.

​

Improved

• Streaming chat is more reliable — fixed a class of stream hangs and “NOSCRIPT” circuit-breaker errors.
• Tool validation errors now include schema context so the LLM can self-correct on the next turn.
• Cleaner log output across the backend.
• Feature flags now auto-set sensible default privileges when toggled on.
• Connector list now scrolls inside its panel instead of moving the whole page.

​

Fixed

• Clicking a starter prompt now puts the text in the input box instead of dropping it.
• SuperAgent uses short, readable step labels for mutating tools instead of full resource names.
• Connector cards no longer stay stuck in “connecting…” when OAuth initiation fails.
• The Disconnect menu on a connector card is no longer clipped by the parent.
• Streaming markdown render path now applies the correct chat styling.

---

​

v0.9.0 — SuperAgent drawer, Skills templates, and a new credit system

​

Added

• SuperAgent redesigned as a right-hand drawer you can open from anywhere in the app.
• Skills templates: kickstart custom skills from a curated library instead of writing them from scratch.
• Skills are now scoped to where you use them — separate sets for New Chat, Agents, and Projects.
• Credit system: see token usage per chat and tenant on the new Token Usage page.
• Embedding model upgraded to model2vec (512-dim vectors) for faster, lighter retrieval.
• API Access section now appears on the agent-create page, not just on existing agents.

​

Improved

• Infisical secrets auto-refresh their token on expiry — no more mid-session auth blips.
• Sandbox acquisition is cluster-aware and refreshes installed skills when the sandbox wakes up.
• Invitation verification no longer crashes on unsupported locales.

​

Fixed

• Chat input shows a spinner on the file chip while a cancel or delete is in flight.
• Deleting a skill keeps the remaining cards visible.
• Workflow Builder now posts token usage directly to the main service.

---

​

v0.8.0 — Multi-tenant home, 30+ new connectors, and SuperAgent v2

​

Added

• Brand-new home page — a dashboard with agent orchestration front and center.
• 30+ new connectors and a tighter category structure to make finding the right one faster.
• Tenant registration service + multi-tenant super admin flow, including a tenant dashboard and hard-deletion worker.
• SuperAgent v2: UX polish, integrated web search, and noticeably smoother streaming.
• Full-fidelity export and import for chat agents — move conversations between environments without losing tool calls or attachments.
• PACER connector with automatic username/password authentication.
• Real-time streaming voice transcription powered by faster-whisper, replacing browser speech recognition.
• Agent usage analytics metrics plus a ready-to-import Grafana dashboard.
• Document classification with usefulness scoring, Australian PII detection, and per-file detail breakdown.
• Excel binary (.xlsb) file uploads.
• HTML artifact generation for agents that need to produce rendered reports.

​

Improved

• Nginx now ships with a stricter security-header set out of the box.
• Express mode is now the default extraction mode for all file uploads.
• File upload cancellation is faster and more reliable on slow networks.
• Browser mic now releases properly when you stop dictating.

​

Fixed

• HITL modal state now persists across chat tab navigation.
• Disabled OAuth-only legal connectors that have no public API.
• Reasoning and tool-call rendering inconsistencies in long chats.