Changelog
Product Updates
New features, improvements, and fixes from the Bothive team. We ship updates weekly to make your workforce smarter.
Today
v6.33.0Page-Aware Hivemind for Mintlify
Bothive documentation assistants can now understand the page a reader is viewing, with a refined Ask Hivemind panel and a subtle scroll-to-ask prompt.
- Added current-page URL, path, title, and bounded Markdown context to Mintlify assistant requests.
- Added a clean page-context indicator, real Bothive logo, responsive theme styling, and clearer assistant loading states.
- Added an Ask Hivemind prompt that appears as readers move through long documentation pages and opens the same assistant panel.
- Added a Mintlify deployment guide covering page context, safe browser tokens, and the native Assistant tradeoff.
Today
v6.32.0Enterprise Teams, Safer Integrations, and Customer Metering
Organisation teams now share agents under stricter permissions, generated integrations run behind stronger network boundaries, and deployments expose real customer usage.
- Centralized organisation access so owners, admins, members, and viewers receive consistent team permissions across every team route.
- Added owner-safe row-level policies for team membership and shared bot assignment.
- Blocked hardcoded secrets, internal-network targets, unsigned webhooks, and unsafe agent definitions before shared HiveLang changes can be staged.
- Pinned custom integration calls to their declared API host, removed token previews from logs, and added response, timeout, redirect, header, and definition limits.
- Added deployed-agent customer plans, subscription entitlements, and real 30-day user, run, subscriber, and platform-cost analytics.
Today
v6.31.0Production Readiness Gates for Agents
Bot overview now shows whether an agent is actually ready for production, and deployments are blocked when critical trust gates fail.
- Added a production-readiness audit covering HiveLang source, deployment records, identity, runtime evidence, observability, action safety, schedules, channels, and memory.
- Added a readiness card to the bot overview with score, blockers, warnings, and direct links to fix weak areas.
- Dashboard deployments now check readiness before activating production so untested or incomplete agents do not publish silently.
- The audit makes the existing enterprise foundation visible: versioned deployments, Pulse schedules, runtime logs, channel/API surfaces, identity memory, and action approval metadata.
Today
v6.30.0OpenAPI Uploads for HiveMind Integration Building
HiveMind can now accept OpenAPI/Swagger JSON specs and use them to generate custom API integrations that bots can call as tools.
- Added a HiveMind OpenAPI upload endpoint that validates and parses specs before generation.
- The HiveMind composer plus button now accepts OpenAPI/Swagger JSON and hands parsed operations to Architect.
- Uploaded specs show service name, base URL, operation count, truncation notes, and parser warnings where relevant.
- Generated integration flows still go through the normal review/connect path so credentials and bot capabilities stay explicit.
Today
v6.29.0PDF Knowledge Uploads and ElevenLabs Agent Voices
Agents can now ingest PDF knowledge safely and speak with configurable ElevenLabs voices from the agent console.
- Hardened knowledge uploads with ownership checks, file type validation, 15MB limits, better PDF metadata, and clearer upload errors.
- Added bot-level ElevenLabs voice profiles with enable/disable state, model, voice, stability, similarity, and speaker boost settings.
- Added voice preview controls in the Identity Studio and playable assistant responses in the bot playground.
- Made ElevenLabs visible as a connectable integration now that the voice experience has product UI.
Today
v6.28.0Gmail Drafts and Clearer Agent Schedules
Agents can now analyze Gmail, leave reviewable drafts, and explain connected integration capabilities while recurring schedules are easier to configure.
- Added gmail.createDraft for daily mail reports and meeting follow-ups that should be reviewed before sending.
- HiveMind now describes connected Gmail, Calendar, Slack, and other integration capabilities in account context.
- Added readable schedule presets and custom cron guidance in the bot builder.
- Added custom Bothive guides for Gmail agents, Slack, and the web widget.
Today
v6.27.0Identity Evaluation Gates
Agent identity changes are now tested against real communication and safety pressure before they can become production policy.
- Added a 20-scenario identity suite spanning chat, email, social, and internal work.
- Evaluations score voice, mission, formatting, boundaries, prohibited claims, and prompt-injection resistance with evidence per scenario.
- Production promotion requires an 80% score, complete scenario coverage, and zero critical violations.
- Changing a draft invalidates its evaluation, preventing stale results from approving different identity rules.
- Evaluation evidence, model usage, weighted credits, and scores are stored with each immutable identity version.
Today
v6.26.0Autonomous Commitment Follow-ups
Conversation promises can now wake their owning agent at the right time, resume safely after failure, and remain visible until the promised outcome is verified.
- Dated commitments now create one-shot Supabase Pulse jobs automatically.
- Follow-up retries use bounded exponential backoff and resume the same durable run to prevent duplicate external actions.
- Agents receive the original commitment, notes, session context, and explicit verification requirements when the timer fires.
- Autopilot now distinguishes scheduled, running, retrying, blocked, failed, awaiting-review, and completed follow-ups.
- Owners can inspect the follow-up trace, run failed work again deliberately, or mark a verified promise complete.
Today
v6.25.0Controlled Agent Learning
Bot memory now detects repeated or contradictory learning before it can change production behavior, keeping owner-approved context trustworthy over time.
- Added shared duplicate and contradiction analysis for manual memories and agent-proposed learning.
- Duplicate proposals are suppressed when approved memory already covers the same context.
- Conflicting proposals remain pending and show the approved memory they disagree with.
- Owners can explicitly replace the previous memory, keep both with an audit marker, or reject the proposal.
- Fixed proposal keys so pending agent memories can be approved through Memory Studio reliably.
Today
v6.24.0Voice and Identity Studio
Production agents now have an owner-controlled identity contract for their role, mission, voice, boundaries, channel behavior, and success criteria.
- Added a dedicated Identity section to every agent console with structured voice, formatting, vocabulary, guardrail, and outcome controls.
- Added channel-specific previews for chat, email, social, and internal communication before promotion.
- Every save creates an immutable identity version with notes and one-click restoration.
- Approved identity contracts are injected into direct, channel, and scheduled runs as owner policy that end-user messages cannot override.
- Existing builder voice profiles and system prompts are carried into the Studio as a starting draft.
Today
v6.23.0Durable Agent Commitments
Entrepreneur agents can now turn promised follow-ups into bot-scoped operational records with due dates, provenance, and an explicit completion state.
- Added authenticated bot-scoped commitment creation and listing.
- Added safe open/done status updates for follow-up ownership.
- Linked each commitment to its bot and originating agent context through metadata.
- Added the commitment.create runtime tool so agents can record concrete follow-ups during execution.
- Added commitment.list guidance so agents can resume existing follow-ups instead of duplicating them.
- Added commitment.complete so verified work can close the exact follow-up that created it.
- Added review states for durable memories; pending learning is excluded from direct, scheduled, and channel runtime recall until approved.
- Added memory.propose so agents can submit candidate learning for owner review without silently changing behavior.
- Added the Autopilot Daily Brief summarizing approvals, recovery work, open follow-ups, and active schedules in one owner view.
- Added deterministic outcome evaluation signals for response quality, criteria completion, execution success, and next action.
Today
v6.22.0Reusable Agent Voice Profiles
Every bot can now start with durable voice and working context, so tone, audience, policies, and preferred formats carry from creation into future runs.
- Added an optional voice and working context field to the builder Configure step.
- Persisted profiles as bot-scoped voice memories after deployment.
- Made direct, scheduled, and channel runtimes recall voice profiles with corrections and verified lessons.
Today
v6.21.0Feedback-Aware Bot Memory
Explicit user corrections can now become durable, bot-scoped learning signals so agents improve from real feedback instead of only successful execution traces.
- Added bot ownership validation before feedback becomes memory.
- Stored explicit corrections with provenance, feedback type, and durable memory metadata.
- Repeated feedback for the same interaction updates one memory instead of duplicating it.
Today
v6.20.0Entrepreneur Agent Intelligence Loop
Bots now run with an outcome-focused intelligence loop that plans work, verifies meaningful actions, records durable lessons, and escalates blockers with useful context.
- Added Perceive, Plan, Act, Verify, Learn, and Escalate runtime guidance for entrepreneurial agents.
- Added persisted intelligence checkpoints for bot runs, including success criteria and learned lessons.
- Resumed runs now continue from their stored intelligence state instead of starting over.
- Added runtime observability for intelligence runs and checkpoints.
Today
v6.19.0Google Meet Runtime Connector
Google Meet is now registered as a first-class Bothive integration so bots can create meeting spaces and work with conference transcripts through the runtime.
- Added Meet to the user-facing integration catalog and OAuth credential resolution.
- Registered meeting creation, conference lookup, transcript listing, and transcript reading as bot capabilities.
Today
v6.18.0Integration Builder Refresh
Creating a custom integration now has a clearer three-step flow with live tool review and HiveMind-generated contracts carried into the connection form.
- Added visible Describe, Add tools, and Test & connect progress framing.
- Added a live HiveLang contract preview before connecting an API.
- HiveMind Architect API research now pre-fills the custom integration form for review.
Today
v6.17.0Google Meet Foundations
Bothive now has the Google Meet connector foundation for creating meeting spaces and retrieving post-meeting transcripts for agent summaries and memory workflows.
- Added Meet OAuth scopes and encrypted connected-account support.
- Added Meet space, conference, transcript, and transcript-entry helpers and HiveLang tool schemas.
Today
v6.16.0Flash Announcements
Admins can now send feature launches and important alerts with an optional image to every user's dashboard and inbox.
- Added an admin announcement composer with title, message, image, and action link fields.
- Added dismissible dashboard alerts backed by the notifications system.
- Added branded email delivery through Resend.
Today
v6.15.0Work-Ready Bot Templates
The builder now includes practical starting points for customer support, FAQ knowledge, and daily job searching.
- Added a knowledge-grounded customer support agent with escalation behavior.
- Added an FAQ assistant for documentation-based self-service answers.
- Added a daily job-search scout that creates a focused shortlist for review.
Today
v6.14.0Scheduled Bot Templates
The builder now includes practical schedule-ready templates for uptime monitoring, lead follow-up, and daily operations briefings.
- Added minute, hourly, and weekday-daily starter schedules.
- The Configure workspace now scrolls correctly on smaller screens so all controls remain reachable.
Today
v6.13.0HiveMind Code Partner
HiveMind can now review and apply targeted improvements to HiveLang directly from the builder Configure panel, with parser validation protecting the editor from broken edits.
- Added compact Recommend and Apply edit actions for the current HiveLang source.
- Existing-bot edit requests now perform a real validated code edit after loading the bot instead of only navigating to the builder.
Today
v6.12.0Supabase Cron Heartbeat
Bothive scheduled agents can now run from a low-cost Supabase Cron heartbeat while Bothive retains job ownership, locking, retries, and execution history.
- Added a Vault-backed Supabase Cron setup for one secure heartbeat per project instead of one provider job per bot.
- Added deployment documentation for configuring, verifying, and removing the heartbeat.
Today
v6.11.0Reliable Scheduled Agents
Scheduled Bothive agents now expose cron failures clearly instead of reporting a healthy run when a schedule or worker has failed.
- Invalid HiveLang schedule code now stops sync with an actionable error while preserving existing schedules.
- The central cron endpoint now returns failure status when Pulse, workflows, action workers, or webhook delivery report errors.
Today
v6.10.0HiveMind Existing Bot Editing
HiveMind can now send owners directly into the builder with an existing bot loaded for updates, fixes, and improvements.
- Edit, update, improve, and fix requests now route to the existing bot's builder workspace.
- The builder preserves the current bot identity and HiveLang source while the owner makes changes.
Today
v6.9.0Community Threads & Builder Responsiveness
Bothive community discussions now support real threaded replies, while the builder is usable on smaller screens and looks correct in light mode.
- Added persistent community comments with reply-ready parent threading and synchronized post counts.
- Added an inline discussion composer and responsive community filters for mobile users.
- Responsive builder panes now stack on mobile and configuration surfaces respect the active theme.
Today
v6.7.0Responsive Dashboard Navigation
The Bothive dashboard now has a focused mobile navigation experience and responsive bot workspaces that stay usable on smaller screens.
- Added a mobile-only dashboard navbar with a theme-aware Bothive logo and no duplicate page header.
- Removed the unreliable mobile sidebar renderer and kept desktop navigation isolated to desktop layouts.
- Improved bot console sizing, wrapping, and overflow behavior for responsive dashboard use.
Today
v6.6.0HiveMind Build Verification & Bot Control
HiveMind-created bots now pass a real HiveLang validation checkpoint before the builder advances, and owners have a clearer way to revisit or remove agents.
- HiveMind drafts are parsed server-side before testing or publishing, with actionable parser errors shown in the builder.
- Successful HiveMind builds now return to the created bot's page so the agent can be immediately revisited and operated.
- The Agents list now includes a confirmation-protected Delete agent action.
- The normal Bothive Connect a bot docs now explain the production API, SDK, browser widget, identity, sessions, and run traces.
Today
v6.5.0Connect Bothive to Your App
Built agents can now move clearly from the Bothive dashboard into a real SaaS through a server API, SDK, or browser-safe widget deployment.
- The Deploy tab now includes a guided Connect Bothive to your app panel with backend, SDK, and widget paths.
- The official SDK run method now uses the public metered /api/v1/chat endpoint instead of an owner-only dashboard route.
- The production widget now uses bothive.cloud by default and sends browser-safe widget tokens to the web channel endpoint.
- Deployment docs now explain API key storage, bot IDs, customer identity, sessions, run IDs, and the server/browser security boundary.
Today
v6.4.0Agent Autopilot Inbox
Bots now have an owner-facing control surface for work that needs approval, recovery, scheduling, or memory review.
- New Autopilot tab shows pending approvals, failed actions, active schedules, and recent learned context in one place.
- Existing runtime traces, action runs, schedules, and memory are unified into a practical daily operating view.
- Bot names are now unique per owner, so repeated templates automatically receive a clear numeric suffix instead of colliding.
- Memory Studio now uses the shared Bothive and shadcn component system for consistent controls and outline actions.
Today
v6.3.0Agent Memory Studio
Agents now have an owner-facing memory layer for building virtual assistants that learn voice, preferences, policies, goals, and reusable context over time.
- New Memory Studio tab inside each agent console for viewing, searching, adding, editing, and deleting long-term bot memories.
- Owner-authenticated memory APIs now expose curated bot memory without creating a separate storage system.
- Memory entries support scopes, importance, tags, and automatic categories such as voice, audience, policy, preference, goal, and fact.
- Voice profiles are highlighted so social and assistant bots can draft in the user's style while still keeping humans in control.
Today
v6.2.0Premium Model Metering & Rich Agent Testing
HiveMind and agent testing now feel more production-ready: premium models consume credits at weighted rates, rich responses render cleanly, and the bot playground has a focused test-lab layout.
- Premium HiveMind model picks such as Gemini Pro, Claude, GPT-5.5, and OpenRouter flagships now drain credits faster based on model cost weight.
- HiveMind answers now render fenced HiveLang/code blocks as polished code panels.
- Markdown-style tables in HiveMind and playground responses render as clean, scrollable tables.
- The bot text playground now uses a two-pane test lab with runtime context, quick prompts, traces, and a cleaner conversation surface.
Earlier today
v6.1.0Workflows, Credits & Smarter HiveMind
A big update to the automation layer: HiveMind now builds runnable workflows from a sentence, AI usage is metered in transparent credits, and scheduled workflows fire on their own.
- HiveMind builds workflows from plain language — and shows them as a card you can open in the visual builder.
- Workflows can chain your existing bots as real Bot nodes — a true swarm, not generic AI steps.
- Scheduled (cron) triggers now run automatically, plan-rate-limited per owner.
- AI usage is billed in credits (1 credit = 1K tokens), pooled across HiveMind, bots, and workflows — see it on the Usage page.
- Inline 'Connect' buttons in HiveMind when a build needs an integration you haven't linked yet.
- New docs: a complete guide to building and scheduling workflows.
Earlier
v6.0.0Bothive Phase 2: The Next Evolution
We are thrilled to announce Bothive Phase 2. This major update brings a suite of powerful new capabilities designed to elevate your AI workforce to unprecedented levels of autonomy and intelligence.
- Architect Mode: Design complex systems visually.
- Swarm Mode: Deploy multiple agents working in unison.
- Prompt Mode: Advanced prompt engineering environment.
- Improved Hivemind AI: Faster, smarter, and more contextual.
- Bot Backend Hosting: Deploy and scale your bots seamlessly.
February 18, 2026
v5.2.1Experience Polish & Stability
A focused update improving the onboarding experience, expanding integration support, and resolving critical authentication flows.
- Onboarding Templates now use HiveLang v5 declarative syntax.
- Added documentation for Gmail, Calendar, and WhatsApp integrations.
- Fixed Google Sign-In redirect loop for seamless authentication.
- Enhanced documentation coverage and navigation.
February 4, 2026
v5.2.0Global Visual Refresh
A complete overhaul of the dashboard aesthetic to align with our premium vision. The dark, glassmorphic theme is now consistent across every pixel of the platform.
- Introduced global ambient background system.
- Refined typography and gradients with theme switching.
- Dynamic, context-aware greetings that change based on time and activity.
- Unified transparency across Sidebar, Legal, and Pricing pages.
January 28, 2026
v5.0.0HiveLang v5: AI-First Syntax
We've reimagined how agents are defined. HiveLang v5 introduces a curly-brace syntax that is optimized for LLM generation and human readability.
- New curly-brace block syntax for Bots and Agents.
- Removed strict JSON requirements for tool definitions.
- Native parallel execution blocks for concurrent agent actions.
- Integrated instructions block for direct LLM system prompting.
January 15, 2026
v4.8.0Marketplace Launch
Monetize your autonomous agents. The Hive Store is now open for creators to publish and sell their specialized assistants.
- Publisher dashboard with earnings analytics.
- Stripe Connect integration for global payouts.
- Verified Agent badges for trusted developers.
- One-click deployment for purchased agents.
December 20, 2025
Enterprise Security & SSO
Hardening the platform for large-scale deployments. We've added robust security features to support our growing enterprise customer base.
- SAML 2.0 / OIDC Support (Okta, Azure AD).
- Granular permission scopes for Bot API keys.
- Audit logs for all agent executions and data access.