Changelog

Every shipped feature, in order.

No stealth launches. No “coming soon”. If it's listed here, it works today. Major versions = new capability. Feature = new surface. Improvement = polish. Fix = bug.

v2.4.0·2026-09-28majorBusiness Brain

Business Brain — an AI team that watches your data and runs the plan

Your agents no longer wait to be asked. Every hour they read your orders, invoices, expenses and books, notice what changed, plan the next step, do the work as drafts and measure what it changed.

›Your plan: one screen with what we noticed, what needs you, what the team is doing and what it finished — with the number before and after each step
›Data watch: orders well below your usual week, a store that went quiet, no invoice for longer than your normal rhythm, expenses or books not updated — every alert carries the exact numbers it is based on
›Playbooks: each alert gets one next step with an owner (Marketing, Sales, Operations or you); if a step does not help, the plan moves to the next one by itself
›Nothing leaves your business on its own: agents prepare drafts, and anything that reaches a customer or moves money waits for your approval
›Approving a weekly priority from your Chief of Staff now actually puts the agent to work on it

v2.3.1·2026-09-28feature

Shopify and WooCommerce orders arrive by themselves

Connect your store in about two minutes, without a developer. Every order is signed by the shop and checked before it counts.

›Integrations → Stores: a personal webhook address, step-by-step setup and a live "last order" status
›Cancellations, refunds and payments update the same order; a late or repeated delivery never overwrites a newer state
›Invoices: record a payment in one click; the invoice closes, reminders stop, and your agents learn how that client pays

v2.3.0·2026-09-27majorCash Intelligence

Cash Intelligence — the Money agent now works on when cash arrives

Five finance capabilities that act, not just report. Each one learns from your own business: client payment habits, forecast accuracy and which alerts you mark as expected.

›Collections agent: predicts when each client will pay, drafts reminders in your voice with the real invoice amount and pay details, and sends only what you approve. Autopilot for early reminders is on Growth+
›13-week cash forecast with an honest low/high band, the lowest week ahead, and "safe to pay yourself today"
›Tax & VAT calendar: deadlines, a set-aside amount and VAT threshold tracking for DE / NL / UK, with every rule linked to its official source
›Credit readiness (Growth+): a lender-style checklist with the concrete steps that move it, plus financing programmes you may qualify for. It is not a credit score
›Spend anomaly alerts (Growth+) and cited industry benchmarks (Pro)
›Guardrails: every figure is checked against your ledger, nothing is sent without approval, and statutory interest is only ever proposed

v2.2.0·2026-09-26majorLead Finder

Lead Finder — describe who you want, get ranked leads

Type what you are looking for in plain language ("200 cosmetics shops in Germany") and get leads ranked for your business from our own crawled pool, never invented.

›AI search: plain-language requests become exact filters for country, industry, city and keywords
›Click the eye to reveal a contact. Monthly unlocks: Solo 3,000, Growth 5,000, Pro 10,000. Extra packs from €19 via Stripe
›Clear email labels: Verified, Catch-all, Verified domain (live site with working mail) and Unverified
›Outreach rules per country are built in. Where cold email is restricted you can still send after confirming you take responsibility
›Send leads straight to chat and the agent drafts the campaign

v2.1.0·2026-09-26featureSequences

Email sequences in the dashboard and in agent chat

Multi-step follow-up sequences now have a Sequences page, and the Marketing agent can create, enrol, pause and report on them for you.

›The agent creates sequences as drafts. Nothing sends until you activate one
›Leads enrolled into a running sequence start paused, so you stay in control
›Every send path checks unsubscribes, bounces and country outreach rules

v2.0.1·2026-09-27improvement

Bring your own model, hardened

Connect any OpenAI-compatible server (Ollama, vLLM, LM Studio, llama.cpp, TGI) at any model size, with a model per task class and automatic cloud fallback.

›Endpoints must be reachable over HTTPS. Use a secure tunnel for a machine in your office
›Private and internal network addresses are refused, and redirects are never followed

v2.0.0·2026-04-14majorPlan 08 / v2 milestone

Automation templates, dry-run sandbox, per-node retry

Visual automation builder reaches production readiness: 8 starter templates, dry-run test mode, and per-node exponential backoff so workflows survive transient failures.

›8 curated starter templates: lead follow-up, invoice recovery, weekly briefing, campaign debrief, onboarding drip, review request, contract chase, re-engagement
›Dry-run mode: every automation can be tested end-to-end without triggering real sends or agent actions
›Per-node retry with configurable exponential backoff (delay, multiplier, max attempts)
›Real wait nodes: pause execution for a fixed duration and resume automatically via BullMQ
›Variable inspector: see resolved values at each node during a dry run
›Template gallery replaces blank canvas as the default new-automation screen

v1.29.0·2026-04-14majorPlan 08

Visual automation builder — canvas and node engine

Draw cross-agent workflows on a React Flow canvas. Connect triggers, agent tasks, wait nodes, conditions, webhooks, and approvals into executable DAGs.

›Drag-and-drop React Flow canvas with auto-layout
›8 node types: trigger (event/schedule), agent task, wait, condition, webhook, email, approval, transform
›Cycle detection at save time — invalid graphs are rejected before they can run
›BullMQ workers execute the graph: wait-node resumes, schedule triggers, retry queues
›Variable templating across nodes: {{trigger.*}}, {{nodes.<id>.*}}, {{run.*}}
›Run history table with per-node status and output values

v1.28.0·2026-04-14featurePlan 07

Agent versioning and prompt quality guard

Every save of a custom agent creates an immutable version. One-click rollback to any prior version. Live prompt quality guard flags issues before you save.

›Semantic version history stored per agent — view diff between any two versions
›Rollback: restore any past version as the active agent in one click
›Prompt quality guard runs before save: flags vague scopes, missing constraints, prompt-injection risks
›Quality guard shows specific, actionable suggestions — not generic warnings
›Version comparison view: old vs new system prompt side-by-side with highlighted diffs

v1.27.0·2026-04-14featurePlan 07

Custom agent builder with 6 starter templates

Build your own agents (HR Coordinator, CS Agent, Finance Analyst) with an 8-step wizard. Six ready-to-clone templates get you to a working agent in minutes.

›8-step creation wizard: name, role prompt, tool allowlist, KB scope, MCP plugins, handoff graph, test console, publish
›6 starter templates: CS Agent, HR Coordinator, Finance Analyst, Content Researcher, Compliance Reviewer, Onboarding Specialist
›Test playground inside the wizard — run the agent against sample inputs before publishing
›Agent analytics tab: task count per day, tool usage breakdown, recent activity log
›Custom agents inherit BaseAgent permission gates, decision trails, and approval flows

v1.26.0·2026-04-14majorPlan 06

Security control library with automatic evidence

A library of security controls mapped to the SOC 2 trust criteria, with automatic evidence collectors and an export for an auditor. It is a tool for preparing an audit — not a SOC 2 report or a certification.

›Controls mapped to the SOC 2 trust criteria (CC1–CC9)
›Automatic evidence collectors: access logs, MFA status, encryption config, API key rotation, etc.
›Audit-readiness score with per-control pass/fail/manual status
›One-click auditor export: machine-readable JSON covering all controls and their evidence
›Per-control notes field for auditor comments and remediation tracking
›/.well-known/security.txt, DPA template, and sub-processor list auto-generated

v1.25.0·2026-04-14featurePlan 06

GDPR data export and 30-day soft deletion

Full GDPR data rights: export everything as a ZIP, request deletion with a 30-day cancellable grace window, and configure per-business retention policies.

›Data export as ZIP with 5 selectable scopes: full, leads, emails, documents, campaigns
›30-day soft-delete: data is flagged, email confirmation sent, permanently deleted after grace period
›Deletion cancellation: owner can cancel within the 30-day window
›Audit retention tiering: 30 days on Solo, 1 year on Team, 7 years on Enterprise
›Per-business data residency declaration for Enterprise DPAs

v1.22.0·2026-04-14featurePlan 04 Phase B

Automated security scanner for MCP plugins

An automated 5-check security scanner runs on every plugin submission and gives the reviewer a verdict — pass, warn or block — with its findings. A person approves or rejects.

›Security scanner (5 checks): manifest shape, Ed25519 signature, command (shell-injection patterns and suspicious flags), URL (https only; localhost and private-network addresses blocked, a raw IP address flagged) and npm package (looked up on the public npm registry; a package that is not a known MCP server is flagged)
›The admin reviewer sees the verdict and every finding next to the manifest in the review dashboard at /admin/marketplace-review
›People browsing a plugin see whether its scan found anything, and how many notes it left
›5-star rating system with verified-install gate — only users who actually installed can rate

v1.21.0·2026-04-14majorPlan 04 Phase A (user side)

MCP plugin marketplace — install, permissions, ratings

End-user side of the MCP marketplace: browse, install, and rate plugins. Permission disclosure modal shows exactly what each plugin can access before install.

›Plugin catalog with search, category filter, and sort by rating / installs / date
›Permission disclosure modal: plain-English breakdown of every scope a plugin requests
›One-click install flow with per-agent allowlist configuration
›Credential management: store and rotate plugin API keys encrypted at rest
›5-star ratings with verified-install enforcement — no fake reviews
›Install count and average rating displayed on every card

v1.20.0·2026-04-14featurePlan 20

Slack integration

@mentions, slash commands, interactive approval cards, channel routing per agent — all the Telegram surface, now in Slack.

›OAuth install flow with full workspace verification
›@Rigyx mentions route to the correct agent
›/rigyx slash commands (inbox, briefing, marketing/sales/ops/pm)
›Block Kit approval cards fork from Telegram (same handlers, zero duplication)
›Daily briefings post to configured channels
›HMAC signature verification on every inbound webhook

v1.19.0·2026-04-14majorPlan 19

Visual automation builder

Zapier-style DAG editor for cross-agent workflows. 13 node types, variable templating, cycle detection at save time.

›Drag-drop canvas powered by React Flow
›13 node types: event/schedule triggers, agent tasks, wait, condition, webhook, email, approval, transform
›Condition nodes use jsonata (no arbitrary JS eval)
›BullMQ workers for wait resumes + schedule triggers
›Variable templating: {{trigger.*}}, {{nodes.*}}, {{run.*}}
›Cycle detection rejects invalid graphs at save
›Seeded example: "Lead qualified → Sales follow-up"

v1.18.0·2026-04-14featurePlan 18

Custom agent builder

Build your own 5th, 6th, Nth agent. Name, role prompt, tool allowlist, knowledge scope, MCP plugins, handoff graph, test playground.

›8-step creation wizard with test console
›Custom agents inherit BaseAgent — permission gates, decision trails, approval flows
›Reserved-slug + dangerous-tool stripping + prompt-injection sanitizer
›Per-business slug prefixes in the agent router
›RBAC extended: custom_agents:read (viewer+), custom_agents:write (manager+)

v1.17.0·2026-04-14majorPlan 17

Compliance tools — audit log, export, deletion, SSO

Audit log, data export, deletion with a 30-day grace period, single sign-on and the DPA documentation.

›Compliance audit log with 20 instrumented sensitive actions
›Full data export as ZIP (5 scopes: full, leads, emails, documents, campaigns)
›30-day soft-delete with email confirmation + cancellation grace window
›SSO: SAML 2.0 + OpenID Connect
›Legal pages: DPA, sub-processors, data retention policy
›Per-business SSO configuration for multi-tenant setups

v1.16.0·2026-04-14featurePlan 16

Embeddable demo widget

Drop a branded Rigyx agent on your customers' sites as a lead magnet. 2.3 KB gzipped loader.

›<script> loader + iframe chat UI with strict CSP
›Origin verification (Origin + Referer headers)
›Per-visitor cookie rate limits + daily/hourly IP caps
›Lead capture auto-creates leads with source attribution
›Per-widget custom branding (logo, colors, copy)

v1.15.0·2026-04-14featurePlan 15

Industry packs

8 pre-built onboarding packs — SaaS, Agency, Local Service, Real Estate, Consulting, Ecommerce, Healthcare, Legal.

›Each pack pre-fills philosophy, brand voice, target audience
›Seeded content plans tied to real business cycles
›Signal sources pre-configured per vertical
›5-8 custom skills per pack with industry-specific triggers
›Idempotent apply — never overwrites user-edited fields

v1.13.0·2026-04-13majorPlan 13

Team seats + RBAC

Multi-user businesses with 5 roles, email invitations, JWT role claims, and role-aware navigation.

›5 built-in roles: owner, admin, manager, operator, viewer
›Custom scope overrides on top of base roles
›Email invitation flow with 7-day expiring tokens
›JWT carries {memberId, role} for fast auth checks
›/auth/me re-reads role for near-realtime revocation
›assertCan(action) permission helper wrapping every sensitive route

v1.12.0·2026-04-13majorPlan 12 Phase A

MCP plugin marketplace (client mode)

Rigyx is now Model Context Protocol native. Connect GitHub, Stripe, Notion, Linear, Slack, Filesystem — or bring your own MCP server.

›MCP client with stdio + SSE + HTTP transports
›Permission gate: per-agent mechanical enforcement
›Tool adapter converts MCP tools to OpenAI function format
›6 first-party plugins seeded: GitHub, Stripe, Notion, Linear, Slack, Filesystem
›Per-tool audit log with SHA-256 args hashing
›Install modal with agent allowlist + credential management
›Connection pooling + 30s timeouts + admin kill switch

v1.11.0·2026-04-13majorPlan 11

Knowledge base + RAG

Upload PDFs, DOCX, URLs, transcripts. Agents auto-retrieve relevant chunks in every task. pgvector + OpenAI embeddings.

›Semantic chunking (800-1200 tokens, 100 token overlap)
›text-embedding-3-small (1536 dim) — ~$0.01/business/month typical
›4 ingest workers: PDF, DOCX, URL, transcript
›Auto-retrieval injected into agent system prompt (20K cap)
›Per-source visibility chips (which agents can read it)
›Search-as-you-type UI with debounced queries
›query_knowledge universal tool on every agent

v1.10.0·2026-04-13majorPlan 10

Public REST API + webhooks

Full public API v1 with scoped OAuth tokens, HMAC-signed webhooks, 29 event types, OpenAPI spec.

›REST API at /api/public/v1/* with sha256-hashed bearer tokens
›15 scope strings for fine-grained permissions
›29 typed event types across leads, campaigns, documents, invoices, agents
›HMAC-SHA256 signed webhooks with 5-min replay protection
›Exponential backoff retry (30s → 24h) with auto-disable
›OpenAPI spec auto-generated at /v1/openapi.json
›Developer tab in portal Settings with token + webhook management
›Cursor pagination + idempotency keys + 60 req/min rate limit

v1.9.0·2026-04-13featurePlan 09

Platform expansion

Instagram DMs, SEO content brief generator, review response agent, proposal→contract automation.

›Instagram DM drafting with per-post engagement references
›SEO content brief via Gemini grounded search + page understanding
›Review response agent with sentiment classification + voice-matched drafts
›Proposal approval auto-generates contract from stored T&C
›Contract clauses library (payment terms, liability, termination, IP, dispute resolution)

v1.8.0·2026-04-12featurePlan 08

Growth lead magnets

Public marketing audit generator, competitor teardown reports, live demo sandbox on agent pages.

›Anonymous URL → branded PDF audit in 90 seconds
›Competitor teardown (3-5 URLs → gap analysis)
›Live demo sandbox on /agents/* pages (1 question per hour per IP)
›Rate limiting + spam prevention
›Email capture + nurture follow-up

v1.7.0·2026-04-12improvementPlan 07

Handoff gap fixes

Three real-world handoff gaps closed: Sales→Ops deal value, mid-flight campaign alerts, invoice-reply routing.

›Sales close now sends deal value + renewal date to Ops AND Project Manager
›Mid-flight campaign underperformance detection (auto-ping Marketing)
›Invoice reply auto-routes to Operations instead of Sales

v1.6.0·2026-04-12majorPlan 06

Multi-provider LLM routing + BYO keys

Tier-based LLM routing (Budget / Balanced / Premium), per-business BYO API keys, Grok/xAI provider added.

›Task-class routing (micro / standard / heavy / browser)
›Encrypted BYO keys for OpenAI, Claude, Gemini, xAI
›Monthly spend caps with soft-warn + hard-pause
›Provider health monitoring with auto-failover
›Inter-agent mailbox for async collaboration
›4 handoff points wired: Reply→Sales, Sales→Ops, Ops→Sales, Marketing→PM

v1.5.0·2026-04-11improvementPlan 05

UI polish pass

Content Library UI, lead tier filters, agent inbox, 9 empty states, mobile responsive fixes.

›Content Library tab surfaces top-performing emails
›Lead list tier filter tabs (Hot / Warm / Cool / Cold)
›Agent inbox for inter-agent messages
›Empty states with CTAs on all major pages

v1.4.0·2026-04-11majorPlan 04

Voice calling (ElevenLabs + Twilio)

Clone your voice once, agents draft voicemails in your voice, Twilio dispatches after approval.

›ElevenLabs voice cloning from 60s sample
›Twilio outbound calls with <Play> TwiML
›Mechanical rate limits: 1 per lead per 7 days
›Calling hours enforced (9am-6pm local timezone)
›Mandatory disclosure prefix per-language (en/bg/de/hu)
›/voice Telegram command to upload sample

v1.3.0·2026-04-11majorPlan 03

Autonomous ad campaigns

Platform-agnostic ad campaign workflow: site analysis → keyword research → draft → review → autonomous browser dispatch.

›Browser agent walks through Google Ads / Meta / LinkedIn on your account
›Always stops at final review screen — never auto-publishes
›Two-gate launch: confirmed=true AND approved_by_user_at
›Pure-JS regex metric parser (never LLM-parsed)
›Gemini-grounded keyword research (no fabricated volumes)

v1.2.0·2026-04-11majorPlan 02

Document generation system

Real PDF proposals/invoices/reports + XLSX lead exports with anti-hallucination review gate.

›Puppeteer + Handlebars PDF templates
›exceljs XLSX generation with formulas
›Mechanical review gate: send hard-blocks on non-approved status
›Operations Agent tools: generate_document, list, send_email
›Refuses missing fields with specific error messages

v1.1.0·2026-04-11majorPlan 01

Frontend redesign

Full marketing site + dashboard redesign. Dark iris/Fraunces aesthetic, NetworkGrid hero, Remotion video, 42 pages.

›Design system: Fraunces + Inter + JetBrains Mono, obsidian/iris/chlor palette
›Animated NetworkGrid hero with mouse parallax
›Remotion programmatic video on landing page + per-agent pages
›MarketingNav + dashboard shell + 14 dashboard sub-pages redesigned
›Accessibility: focus rings, ARIA, reduced-motion respected

Want something shipped faster?

Every customer request goes on the roadmap. The loudest complaints get fixed first. Tell us what's missing.

[email protected]Start free