# FourIA: Vision & Value Proposition **Product:** FourIA **Version:** v1.0-beta — Design Partner Program **Beta Period:** July 16, 2026 — Open-ended (GA date to be determined) **Owner:** Paso4 **Date:** July 16, 2026 --- ## One-liner FourIA is a managed, private AI agent runtime — bring your markdown-based agents and give them a body to operate, integrate with your infrastructure, and serve your clients. --- ## What is FourIA? FourIA is an **agent orchestration runtime** that turns markdown-based AI agents into operational systems running on private infrastructure. Businesses use it to deploy domain-specific agents that serve their clients — securely, privately, and without managing the underlying platform. Three pillars define the platform: 1. **Fouria Runtime** — The execution layer. Bring agents defined as markdown files and FourIA gives them a body: connections to your data, railguards and tools to validate reasoning, event listening, and scheduled execution. Your agents don't just chat — they operate. 2. **Plug & Play with Private Infrastructure** — Encrypted storage, automated backups, secrets and API token management, centralized security patches, and the choice between commercial or self-hosted AI. A strict No Logs policy combined with Personal Information data masking ensures that AI providers never read your or your clients' data. Privacy is architecture, not policy. 3. **Bring Your Own Lunch (BYOL)** — The system architecture is modular. Customers who provide their own storage or LLM infrastructure receive direct discounts. No vendor lock-in, no mandatory stack. FourIA also offers extra features that extend the core runtime. The **FourIA Digital Twin** — automated knowledge generation through the LLM Wiki pattern — is currently under active development as an extra feature outside the core runtime. See [Extra Features](#extra-features) for details. --- ## The Problem FourIA Solves ### Businesses need domain agents — but don't need the platform AI agents are becoming a competitive necessity. A law firm wants a legal research agent for its clients. A consulting firm wants a market intelligence agent. A real estate developer wants a deal analysis agent. These are **domain-specific agents** — they need tailored knowledge, custom tools, private data, and reliable execution. Building and running these agents in production requires solving problems no business should have to solve: - **Infrastructure management** — Container orchestration, uptime, scaling, backups, disaster recovery - **Security and compliance** — Secrets rotation, API token management, encryption, zero-trust data access - **Privacy architecture** — Ensuring that AI providers never see sensitive client data, implementing PII masking, maintaining a verifiable No Logs posture - **Multi-tenancy** — Running agents for multiple clients with strict isolation guarantees - **Monitoring and billing** — Tracking costs per agent, per client, per action — without logging content FourIA solves all of this as a managed runtime. Businesses focus on designing their agents (the domain expertise). FourIA handles everything else (the platform). ### The Partnership Opportunity Two partnership models serve different needs: 1. **Domain Agents for Clients** — A business deploys domain-specific agents that their own clients interact with. Example: a law firm running legal research agents accessible to their corporate clients. FourIA provides the runtime; the business provides the domain expertise. 2. **Reseller / White-Label** — Agencies, consultancies, and technology partners resell FourIA as a platform under their own brand. They bundle agent design, deployment, and support as a managed service to multiple businesses. For both models, Paso4 offers an **Agentic Consultancy** service — our expert agent designers provide paid professional services to optimize agent performance, reduce token costs, and improve completion rates. --- ## How FourIA Works ``` ┌──────────────────────────────────┐ │ Your Domain Agents │ │ Markdown-based, portable, owned │ └──────────────┬───────────────────┘ │ ▼ ┌──────────────┐ ┌──────────────────────────────────┐ ┌──────────────┐ │ Your Data │───▶│ Fouria Runtime │───▶│ Your Clients │ │ │ │ │ │ │ │ • Documents │ │ Agent Execution │ │ • Chat │ │ • APIs │ │ Event Listening │ │ • Telegram │ │ • Databases │ │ Scheduled Automation │ │ • Discord │ │ • Webhooks │ │ Railguards & Validation │ │ • Slack │ │ │ │ Tool Integration (MCP) │ │ • API │ └──────────────┘ │ │ └──────────────┘ │ ◀── Plug & Play Infrastructure ──│ │ Encrypted Storage, Secrets Mgmt, │ │ Security Patches, No Logs + PII │ │ Masking, BYOL Architecture │ └──────────────────────────────────┘ │ ▼ ┌──────────────────────────────────┐ │ BYOL: Bring Your Own Stack │ │ Storage (R2, S3) │ │ LLM Providers (Anthropic, OpenAI, │ │ self-hosted, AI Gateway) │ │ → Discounts for each BYOL module │ └──────────────────────────────────┘ ``` **Key architectural insight:** FourIA sits between your data and your clients. You define the agents — their behavior, domain knowledge, and tool access — as portable markdown files. FourIA gives them runtime, infrastructure, and safety. Your clients interact with the agents through any supported channel. You own the agents, the data, and the client relationship. --- ## Core Features ### Fouria Runtime The execution engine that turns markdown agent definitions into operational systems. | Capability | Description | | --------------------------- | ------------------------------------------------------------------------------------------------------------------- | | **Agent orchestration** | Deploy multi-agent systems with isolated workspaces, tool access, and session management | | **Data integration** | Connect agents to your documents, APIs, databases, and webhooks — safely sandboxed | | **Railguards & validation** | Built-in tool policies, approval gates, and execution sandboxes so agents validate reasoning before acting | | **Event listening** | Agents respond to incoming messages, webhooks, and system events across channels | | **Scheduled automation** | Program agents to run on cron schedules — wake, execute, report | | **Channel routing** | Deploy agents to Telegram, Discord, Slack, Microsoft Teams, Google Chat, WhatsApp — with per-channel access control | | **Multi-agent routing** | Route messages to specialized agents by channel, sender, or content | Agents are defined as **markdown files** — portable, version-controlled, and owned by you. No proprietary format. No vendor lock-in. ### Plug & Play with Private Infrastructure Managed infrastructure with privacy as a structural guarantee. | Capability | Description | | ---------------------------------- | ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | | **Encrypted storage & backups** | All agent state, memory, and configuration encrypted at rest with automated snapshot backups | | **Secrets & API token management** | Centralized secret store with rotation support — tokens never logged, never exposed to agent output | | **Centralized security patches** | Platform-level updates for the agent runtime, eliminating per-deployment maintenance | | **Commercial or self-hosted AI** | Route through Anthropic, OpenAI, Cloudflare AI Gateway, or your own self-hosted LLM endpoint | | **No Logs policy** | Message content, prompts, responses, and tool outputs are never collected, stored, or transmitted. Only operational metrics (channel type, timestamp, action count) for billing | | **PII data masking** | Automatic masking of personal identifiable information before any external API call — AI providers never read your or your clients' data. Delegated to Cloudflare AI Gateway natively (no separate outbound worker required). | See [FourIA Privacy Policy](/llms/platform/apps/fouria/privacy/index.md) for the complete data handling specification. ### Bring Your Own Lunch (BYOL) Modular architecture means you can bring your own components and pay less. | Module | What you bring | Discount | | ---------------- | ---------------------------------------------------------------- | ----------------------------------------- | | **Storage** | Your own R2 bucket, S3 bucket, or compatible object store | Reduced platform fee | | **LLM provider** | Your own API keys for Anthropic, OpenAI, or self-hosted endpoint | Reduced per-action cost | | **AI Gateway** | Your own Cloudflare AI Gateway account | Direct provider billing, no FourIA markup | Each BYOL module you provide reduces your FourIA bill. The more of your own stack you bring, the less the platform costs. If you prefer an all-inclusive managed experience, we provide everything — at full rate. --- ## Partnership Model FourIA supports two partnership categories. Both benefit from the same core runtime, private infrastructure, and BYOL flexibility. ### Domain Agent Partnerships Businesses that deploy domain-specific agents for their own clients. **Example:** A law firm builds a legal research agent and a contract analysis agent. Their corporate clients interact with these agents through Slack or a web interface. The law firm owns the agent design and client relationships. FourIA runs the platform. **What the partner provides:** - Domain expertise and agent design - Client relationships and support - Domain-specific tools and data sources **What FourIA provides:** - Managed runtime with private infrastructure - Channel connectivity (Telegram, Discord, Slack, etc.) - Security, privacy, and compliance guarantees - Usage billing per completed action ### Reseller / White-Label Partnerships Agencies, consultancies, and technology partners who resell FourIA as a managed platform to multiple businesses under their own brand. > **v1.0:** Ships with ROOT/ADMIN/BASE_USER roles. Partner hierarchy (Partner root role with multi-tenant management) available in v1.1+. **What the partner provides:** - Client acquisition, onboarding, and support - Agent design and configuration services - Bundled pricing and packaging - Brand and go-to-market **What FourIA provides:** - The full runtime platform under the partner's brand - Per-client isolation and multi-tenancy - White-label dashboard and API - Usage-based wholesale billing ### Agentic Consultancy (Professional Services) Paso4 offers paid professional services for partners who want expert guidance: - **Agent design review** — Optimize agent architecture, tool definitions, and prompt structure - **Cost optimization** — Analyze action patterns and LLM usage to reduce per-action cost - **Performance tuning** — Improve completion rates, reduce failure loops, and enhance reasoning quality - **Custom tool development** — Build domain-specific MCP tools and integrations Consultancy is billed separately as professional services — 6 hours/month minimum engagement. It is not bundled with the platform during general availability. Partners engage consultancy on a project or retainer basis. > **v1.0-beta exception:** During the Design Partner Program, agentic consultancy is bundled at 6 hours/month at no extra cost. Post-beta, consultancy returns to the separately-billed model described above. --- ## Extra Features > **Note:** Extra features are not included in FourIA's core runtime functionality. They extend the platform with additional capabilities and are subject to separate availability, pricing plans, and development status. Extra features are billed independently from the core runtime credits. --- ### FourIA Digital Twin (Under Development) > **Status:** Currently under active development. Not available in the core runtime. Will have its own pricing plan, separate from core runtime credits. Release timeline and pricing to be announced. The FourIA Digital Twin is an automated knowledge generation system built on the [LLM Wiki pattern](https://karpathy.ai) (Andrej Karpathy, 2026). It converts fragmented company information into a verified, self-maintaining knowledge graph — a digital twin of how your business actually works. Four core characteristics define how it operates: 1. **Humans provide sources, not answers.** You point the Digital Twin at your documents, Slack threads, financial models, and legal briefs. The system reads, synthesizes, and cross-references them. You validate the results. 2. **Knowledge compounds, it doesn't decay.** Every new source doesn't just get indexed — it integrates into an existing wiki. Cross-references form automatically. Contradictions surface. Stale entries flag themselves. The graph gets richer over time, not noisier. 3. **AI maintains, humans verify.** The tedious work — updating cross-references, flagging contradictions, keeping summaries current, detecting stale information — happens automatically. People approve or reject. The system generates, humans validate. 4. **The output is yours, permanently.** Everything is stored as human-readable Markdown. Portable, auditable, usable without AI. No vendor lock-in. Export everything at any time. Think of it as **a company-wide brain that writes itself** — not a search engine that finds your old documents, but a living model of what your business knows, who knows it, and how it all connects. #### The Problem the Digital Twin Solves **Today's workflow in business knowledge work:** 1. **A decision-maker needs an answer.** They open a week-long ChatGPT or Claude conversation and type a direct prompt. The LLM generates text from scratch, with no memory of the last conversation, no access to internal documents, and no traceability. 2. **The answer is incomplete.** The LLM can't reference last quarter's financial model, the legal opinion from three months ago, or the memo the COO wrote but never shared — because that knowledge lives scattered across OneDrive folders, Slack channels, SharePoint sites, and email threads. 3. **So they interrupt a teammate.** "Hey, do you remember what we decided about the zoning variance?" This question has been asked and answered before. But the answer lives in someone else's head, or buried in a Teams thread from six months ago. 4. **Or they search manually.** Open SharePoint. Search by keyword. Get 47 results, none recent, most irrelevant. Switch to Teams. Scroll through months of history. Find a thread that references the answer but links to a document that was moved. 5. **The bigger companies buy a tool.** Platforms like Knowler connect to a limited set of knowledge bases and let teams search by keyword. The search returns document chunks, not answers. It doesn't synthesize, it retrieves. It doesn't maintain, it indexes. 6. **The even-bigger companies buy a better tool.** Guru integrates multiple knowledge sources, fetches verifiable chunks via RAG, and improves querying through an LLM layer. But Guru organizes what you already know — it does not generate new understanding or cross-reference contradictions across sources. 7. **The largest organizations pay for Palantir's Ontology.** Military and government environments model massive amounts of data into structured graphs. This works, but it costs millions, takes months to implement, and requires dedicated teams to maintain. **The pattern across every stage:** Knowledge that should be an asset remains trapped. In conversations that overflow. In documents that rot. In people who leave. #### With the Digital Twin 1. **Point it at your sources.** Connect OneDrive, Slack, SharePoint, email, internal memos, financial models. The system ingests them automatically. 2. **The wiki writes itself.** The Digital Twin reads each source, extracts key information, creates entity pages, concept pages, and cross-references — updating the existing graph with every new document. Contradictions are flagged. Stale entries generate reports. 3. **You validate, not author.** When it surfaces a contradiction between the Q3 financial model and last week's board memo, you decide which is current. The system maintains; humans approve. 4. **Your digital twin compounds.** Every answer, every correction, every new source makes the graph stronger. Questions produce answers that get filed back in. Nothing is lost to context window overflow. 5. **Query your organization.** "What did we decide about the Algaba zoning variance?" — answered from the verified knowledge graph with sources and confidence levels. **One phrase. Compounding. Verified. Structural.** #### Why Digital Twin Makes Business Better | Without Digital Twin | With Digital Twin | | --------------------------------------------------------------------------------- | ------------------------------------------------------------------------------------------- | | Decisions made from fragments scattered across 5+ tools | Decisions made from a complete, verified knowledge graph | | LLM conversations with no memory between sessions | A persistent knowledge base that compounds with every interaction | | Teammates interrupted for information they've already shared | Teammates consulted only for new judgment, not repeated recall | | Keyword search that returns document chunks out of context | Structural answers that synthesize across sources with provenance | | Knowledge walks out the door when people leave | Knowledge stays, verified, in the organizational digital twin | | Paid analysts re-deriving answers from raw documents every quarter | Analysts querying a wiki where past analysis is already cross-referenced | | Platforms like Guru that organize what you wrote but don't generate understanding | A system that reads your documents, synthesizes them, and builds connections you never made | | Palantir-scale investment required to model organizational knowledge | Self-maintaining knowledge twin at a fraction of the cost | #### Digital Twin Architecture ``` ┌──────────────────────────────────┐ │ People (Validate) │ │ "Yes, this is current." │ │ "No, that contradicts this." │ └──────────────┬───────────────────┘ │ ▼ ┌──────────────┐ ┌──────────────────────────┐ ┌──────────────┐ │ Raw Sources │───▶│ FourIA Digital Twin │◀───│ Queries │ │ │ │ │ │ │ │ • Documents │ │ Ingest ──▶ Synthesize │ │ • "What did │ │ • Slack │ │ Query ──▶ Answer + Cite │ │ we decide │ │ • Email │ │ Lint ──▶ Detect stale │ │ about X?" │ │ • Financials │ │ │ │ • "Show me │ │ • Legal briefs│ │ ◀── Human Verification ──│ │ the trend" │ │ • Memos │ │ │ │ • "What's │ │ • Spreadsheets│ │ ──▶ Compounding Knowledge ─▶│ │ missing?" │ └──────────────┘ └──────────────────────────┘ └──────────────┘ │ ▼ ┌──────────────────────────────────┐ │ Always-Portable Output │ │ Obsidian Markdown + Exports │ │ (Yours permanently, no lock-in) │ └──────────────────────────────────┘ ``` The methodology is the **LLM Wiki pattern**: a three-step workflow where the LLM handles ingestion, synthesis, and maintenance of a persistent wiki, while humans stay in charge of sourcing and verification. The FourIA Digital Twin automates this flow using OpenClaw's Memory Wiki system — so the manual interaction Karpathy's pattern requires happens on its own. #### Digital Twin Example Workflows **Business Intelligence from Existing Knowledge** A real estate investment firm has 18 months of deal analyses, financial models, market reports, and board memos spread across OneDrive, email, and Excel. 1. Connect OneDrive and email. 2. Over two weeks, the Digital Twin ingests sources, creating entity pages for each deal, concept pages for market segments, and synthesis pages that cross-reference financial projections with actual outcomes. 3. The founding partner asks: "Which market segments showed the most deviation between projected and actual returns?" 4. The Digital Twin returns a synthesized answer citing 6 verified sources with confidence levels. 5. The answer is filed back into the wiki as a new synthesis page — this analysis won't need to be re-derived next quarter. 6. When Q4 financials arrive, the system detects that the market segment page needs updating. **Legal Advisory Performance Boost** A legal team has years of contracts, opinions, regulatory filings, and case briefs stored in shared drives and email. 1. Connect the document base and email archives. 2. The Digital Twin creates entity pages for key legal concepts, jurisdiction-specific regulations, and precedent relationships. Contradictions between a 2023 regulatory opinion and a 2025 amendment are flagged automatically. 3. A junior associate asks: "What's our position on the urban development exemption in the Algaba zone?" 4. The Digital Twin provides a cited answer drawing from 4 verified documents. 5. When new case law emerges, the system identifies which entries reference the overruled precedent and surfaces them for review. **Founder Exit: Digital Twin as an Asset** A founder is preparing their company for acquisition. 1. The Digital Twin has been running for 12 months, ingesting internal documents, Slack discussions, financial models, and client correspondence. 2. The digital twin contains entity pages for every key relationship, concept pages for operational processes, synthesis pages for strategic decisions, and a complete audit trail of verification. 3. The founder exports the entire knowledge graph as a deliverable asset. 4. The acquiring team receives a navigable, self-documenting model of how the business works. #### Digital Twin USP **The first system that makes knowledge management possible by eliminating the maintenance cost that kills every wiki.** Existing solutions fall into two categories: - **Retrieval tools** (SharePoint search, Knowler) find documents you already have. They don't synthesize, cross-reference, or maintain anything. - **Organization tools** (Guru, Confluence) structure explicit knowledge someone already wrote down. They improve the query experience but don't generate new understanding. Both assume someone will do the work of writing, updating, and cross-referencing. That someone never does — and wikis die. The Digital Twin's approach is structurally different: 1. **The LLM Wiki methodology** — A proven pattern where the AI handles all bookkeeping: creating pages, updating cross-references, flagging contradictions, maintaining consistency. Humans provide sources and verify. 2. **Automated via OpenClaw's Memory Wiki** — Ingestion, linting, contradiction detection, and synthesis happen automatically through a production-ready plugin with structured claims, evidence, provenance, and confidence scoring. 3. **Self-maintaining digital twin** — The result isn't a wiki that decays. It's a living model of your organization that compounds with every source and every question. #### Digital Twin Analogies | If you know... | The Digital Twin is like... | | ---------------------------------------- | --------------------------------------------------------------------------------------------------------------------------------------------- | | **ChatGPT / Claude** | A version that actually remembers your business — not just the last 200 messages, but everything documented, cross-referenced, and maintained | | **SharePoint / Confluence** | A wiki that writes itself, maintains itself, and flags contradictions instead of letting them rot | | **Guru** | The next step beyond organizing explicit knowledge — generating new understanding from your sources | | **Palantir's Ontology** | A navigable model of your organization's data and processes — self-maintaining, at a fraction of the cost | | **Obsidian / Notion** | A knowledge base where you never manually create pages, update cross-references, or check staleness | | **RAG (Retrieval-Augmented Generation)** | The compounding version: a persistent wiki that gets richer with every source instead of re-discovering from raw documents each query | > **Reminder:** The FourIA Digital Twin is an extra feature, not part of the core runtime. It is currently under active development and not yet available. It will have its own pricing plan, separate from core runtime credits. Release timeline and pricing will be announced separately. --- ## Who Should Use FourIA? ### The Business Operator A company that needs domain-specific AI agents for their operations — legal research, market intelligence, deal analysis, customer support — but doesn't want to build and maintain the infrastructure. They bring their domain expertise. FourIA brings the runtime. ### The Consulting Firm A consultancy that wants to offer AI agent services to their clients under their own brand. They use FourIA's white-label partnership to deploy, manage, and bill — focusing on client relationships and agent design rather than platform engineering. ### The Technology Partner An agency or development shop that builds custom AI solutions. They resell FourIA as the managed infrastructure layer, bundling it with their own agent design, integration, and support services. ### The Domain Expert A professional who understands their field deeply — law, finance, real estate, healthcare — and wants to productize that expertise into an agent that serves clients. They define the agent's knowledge and behavior. FourIA handles everything else. ### The Decision-Making Principal A founding partner, principal, or owner-operator who needs better decisions from complete information. When Digital Twin becomes available, this persona benefits from automated knowledge synthesis. Today, they benefit from agents that give them persistent context instead of ephemeral LLM conversations. ### The Exiting Founder A founder preparing for acquisition who wants to capture organizational knowledge as a transferable asset. When Digital Twin becomes available, they can export a complete, verified knowledge graph. Today, they can deploy agents that document processes and decisions as they happen. --- ## Pricing FourIA pricing is based on **credits** — the exclusive billable unit of the platform. Credits are consumed when an agent completes work. The terms "credit" and "action" are interchangeable throughout FourIA's documentation and billing. > **v1.0-beta:** Delivered through a two-phase Close Beta → Open Beta program designed for small businesses. Pricing and feature scope evolve with each phase. ### Credits (Actions) A credit is consumed when an agent completes a unit of work. Each completed credit is billable: | Credit type | Definition | Credits consumed | | -------------- | ------------------------------------------------------------------ | ---------------- | | **Message** | One inbound message processed by an agent (any channel) | 1 | | **Automation** | One scheduled or event-driven background task executed by an agent | 1 | **In plain language:** Messages are when someone chats with your agent — like asking it a question or giving it an instruction. Automations are background tasks your agent runs on its own — like checking for new data every morning, generating a daily report, or watching for a webhook event. Both consume one credit on successful completion. **Credit ≠ AI expense.** A credit covers the platform execution — infrastructure, orchestration, railguards, and channel delivery. AI provider costs (LLM tokens, model inference) are separate from credits. Customers either pay AI costs directly via BYOL (bring your own API keys) or through FourIA's managed AI Gateway billing. **Success-only billing:** If an action does not complete successfully, the credit is not consumed. You pay for results, not attempts. ### Reader Users vs. Builder Users FourIA distinguishes between two user types: | User type | Capabilities | Pricing | | ----------- | -------------------------------------------------------- | ---------------- | | **Reader** | View agents, channels, and dashboards. Read-only access. | Unlimited, free | | **Builder** | Create/edit agents, configure channels, manage tools | Counted per seat | All users must be registered in the platform. Reader users are unlimited and free on all plans. Builder seats are limited per plan and determine how many team members can design and deploy agents. ### v1.0-beta Plan (Close Beta → Open Beta) The v1.0-beta delivers FourIA through a two-phase program designed for small businesses needing business process optimization. It is not a self-serve plan during the Close Beta window. | Plan attribute | Phase 1: Close Beta (Jul 16 – TBD) | Phase 2: Open Beta (TBD – onward) | | ----------------------- | ---------------------------------------------------- | ------------------------------------------------ | | **Platform fee** | €0 — free | 185€ + VAT / month | | **Credits included** | 2,000 / month (hard cap) | 2,000 / month (overage packs available) | | **Generated memory** | 50 GB (free) | 50 GB (free) | | **AI provider costs** | BYOL only — client provides own LLM keys | BYOL only — client provides own LLM keys | | **Builder seats** | 2 | 3 | | **Reader users** | Unlimited (free) | Unlimited (free) | | **Running agents** | 5 | 10 | | **Channels** | 2 | 3 | | **Integrations** | 5 | Unlimited | | **Login** | GitHub OAuth | GitHub OAuth | | **Data policy** | Staging data only — no real production data | Production data permitted | | **Agentic Consultancy** | Bundled — 2 sessions/month (design + implementation) | Not bundled — available as professional services | **Phase 3 — Post-Beta (GA):** After the Open Beta phase, platform terms will be renegotiated before general availability. V1 GA must include a stable and validated set of functional requirements gathered from multiple clients — confirming that FourIA can operate an agentic runtime according to real-world domain needs. The GA date is not yet determined and will be defined once these requirements are validated. #### Phase 1: Close Beta (Free, Jul 16 – date reviewable) The Close Beta is a co-creation window. Small business clients use FourIA at no platform cost while Paso4 shapes the product around their real needs. **All dates are reviewable if goals are achieved earlier.** **Core principle:** "We will shape the product around you." **Included:** - Model agnostic capabilities (use any LLM provider, BYOL keys required) - GitHub OAuth login - 5 running agents - 5 integrations (APIs, databases, webhooks) - 2 channels (Telegram, Discord, Slack, WhatsApp, MS Teams, Google Chat — any two) - Automatic encrypted backups - Knowledge Wiki generator (Markdown output) - Success-only billing (failed actions consume no credits) - **2 free automation consultancy sessions/month** — design and implementation support by Paso4's agentic engineers - 50 GB of generated memory storage **Data policy during Close Beta:** - **Staging data only** — no real production data until functional validation is complete - Purpose: verify agents work correctly with your workflows using representative test data - Full production onboarding happens at the Open Beta transition once both parties validate readiness **During Close Beta, AI provider costs:** - Client provides their own LLM API keys (BYOL mandatory) - No AI cost coverage from Paso4 #### Phase 2: Open Beta (Paid, date determined by milestone completion) Same features as Close Beta, with expanded limits: - **185€ + VAT / month** for 2,000 credits - 10 running agents (up from 5) - 3 builder seats (up from 2) - Unlimited integrations (up from 5) - 3 channels (up from 2) - Production data permitted after functional validation sign-off - Extra credit packs available on demand without plan upgrade - Agentic Consultancy available as separate professional services **LLM Wiki billing (extra feature, active development):** The FourIA Digital Twin (LLM Wiki) operates on three operations: | LLM Wiki operation | Billing | | ------------------ | --------------------------------------------------------- | | **Ingest** | Billable — charged per document ingested into the wiki | | **Lint** | Billable — charged per cross-reference validation run | | **Query** | Included — consumed as part of message/automation credits | LLM Wiki is an extra feature with independent pricing, not part of the core runtime credit pool. **Not included (available post-beta or on higher plans):** - SSO login (SAML/OIDC) - Group users and agents by groups and channels - Advanced analytics dashboard - Model automatic assignment optimization (best model per scenario) - Knowledge graph visualizer - REST API access - MCP tool access ### Close Beta Program The Close Beta is an engagement with small business clients who want to optimize their operations with FourIA. The program trades hands-on collaboration for real-world validation and the first documented business cases. **Client obligations during Close Beta:** 1. **One documented business case** — Co-authored over the beta window. Becomes FourIA's first published case study (named or anonymized, client's choice). 2. **Weekly sync** — 30-minute structured call. Feedback logged to a shared board — blockers, confusions, and "aha moments." 3. **Staging data validation** — Test workflows with representative data before moving to production. **Paso4 obligations during Close Beta:** - **2 free agentic consultancy sessions/month** — Design review and implementation support at no extra cost during Close Beta. Post-beta, consultancy returns to the separately-billed professional services model. - **Transparent incident communications** — Within 1 hour of detection, via the client's chosen channel. - **Free diagnostics** — Any FourIA-side error or diagnostic submission is not billed. - **Early Adopter perks locked in** — After GA, Close Beta clients retain: locked-in pricing for 12 months, special discount group standing, and grandfathered consideration for any post-GA tier changes. - **Flexible timeline** — If all validation milestones are achieved earlier than projected, the transition to Open Beta can be advanced. Dates are not rigid — they reflect effort estimates, not contractual deadlines. **SLA stance — No SLA during v1.0-beta:** FourIA is operated under **Beta Operational Intent** during the v1.0-beta window. Paso4 commits to transparent incident communications within 1 hour of detection and free diagnostics. No uptime, availability, or data-recovery guarantee is made. R2 backups are best-effort recovery only. Beta status is acknowledged in the client agreement. An SLA will be established once the platform has been validated across multiple clients and failure recovery scenarios. **V1 GA definition:** V1 general availability requires: - A stable, validated set of **functional requirements gathered from multiple clients** — proving FourIA can operate an agentic runtime according to domain-specific business needs, not just technical happy paths - Successful operation with Close Beta clients across both phases - At least one published business case demonstrating measurable efficiency or cost improvement versus an existing tool - Zero data-loss incidents on live tenants - Completion of the 3-workflow benchmark comparison (see Benchmark Methodology below) The GA date is determined by meeting these criteria, not by a calendar deadline. ### Benchmark Methodology FourIA tracks efficiency and cost against alternatives using two complementary approaches: **A. Three reference workflows (client-selected)** Each Close Beta client selects up to 3 workflows from their real business. Each is implemented in FourIA and at least one comparison tool (n8n, Zapier, or Claude Cowork — client's choice per workflow). Three dimensions are measured per workflow: | Dimension | How it's measured | | ------------------------------- | ------------------------------------------------------------------ | | **Cost per successful outcome** | Each tool's billing unit normalized to € per successful completion | | **Time-to-deploy** | Minutes from blank tenant to first successful execution | | **Quality** | Partner rates output on a 1–5 scale per workflow run | **B. Cost-per-successful-outcome metric (FourIA's published KPI)** FourIA's formula: `(platform_credits_cost + AI_gateway_cost) / successful_outcomes` Because FourIA uses success-only billing, this metric is naturally honest — failed actions consume no credits. We publish FourIA's number per benchmark workflow and challenge competitors to publish theirs on the same workflows. Marketing-led; the 3-workflow comparison is the engineering-led counterweight. ### Beta Success Criteria (Internal) Five gates determine whether the beta is ready to conclude and GA planning can begin: 1. At least one documented business case published. 2. Benchmark workflows completed with measured cost, time-to-deploy, and quality on FourIA and at least one comparison tool. 3. NPS ≥ 40 from Close Beta clients at day 90 of the Open Beta phase. 4. Zero data-loss incidents on any live tenant. 5. The 2,000-credit Open Beta phase completed without runaway AI cost surprise or cost-tracking regression. ### Partner Pricing Partners (resellers and white-label operators) pay a **fixed credit fee** discussed directly with the Paso4 sales team. Partner pricing is not self-serve and is negotiated based on volume, client count, and engagement scope. Close Beta and Open Beta pricing above applies to non-partner, single-organization deployments. ### BYOL Discounts Customers who provide their own infrastructure reduce AI-related costs (not credit costs): | Module you provide | Effect | | ----------------------------------------------- | ----------------------------------------------------------- | | Your own storage (R2, S3, or compatible) | Reduced platform infrastructure fee | | Your own LLM API keys (Anthropic, OpenAI, etc.) | Customer pays AI provider directly — no FourIA markup on AI | | Your own self-hosted LLM endpoint | Customer bears inference cost — no FourIA markup on AI | | Your own Cloudflare AI Gateway account | Direct provider billing — no FourIA AI markup | > BYOL discounts affect AI and infrastructure costs, not the credit price. Credits always cost the same regardless of infrastructure choices. ### Billing vs. Cost Tracking FourIA maintains a strict separation between billing and cost observability: | Domain | What it measures | Purpose | | ----------------- | ---------------------------------------------------------------------------------------------------------------------------------------- | ----------------------------------------------------------------------- | | **Billing** | Credits consumed (messages, automations) | Client invoicing, plan limits | | **Cost tracking** | All platform events (AI queries, container activity, R2 operations, cron triggers, WebSocket sessions, action completions, action steps) | Internal cost analysis, infrastructure optimization, margin calculation | Billing events are a subset of cost tracking events. Cost tracking is comprehensive and includes all telemetry from the platform — it is never used for client billing directly. This independence ensures billing remains simple and predictable (credits consumed) while cost data drives internal platform decisions. ### Agentic Consultancy Paso4's agent design and optimization services are billed separately as professional services — 6 hours/month minimum engagement. Rates are project-based or retainer-based, depending on engagement scope. Consultancy is optional and never bundled into the platform price during general availability. > **v1.0-beta exception:** During the Close Beta phase, 2 agentic consultancy sessions/month (design review + implementation support) are bundled at no extra cost. Post-Close Beta, consultancy returns to the separately-billed professional services model. ### Pricing Summary | Question | Answer | | --------------------------- | --------------------------------------------------------------------------------- | | What do I pay for? | Credits — consumed by completed messages and automations | | What about AI costs? | Separate from credits. BYOL (your keys) or managed AI Gateway billing | | What drives the cost? | Credit volume. Reader users are free. Builder seats are plan-limited | | What happens with failures? | Failed actions consume no credits | | Can I reduce my bill? | Yes — bring your own storage, LLM keys, or self-hosted AI | | Is there a minimum? | No. Pay for the plan that fits your credit needs | | Open Beta plan cost? | 185€ + VAT/month for 2,000 credits (extra packs available) | | What's free? | 50 GB of generated memory. Reader users. 2 consultancy sessions during Close Beta | | LLM Wiki cost? | Ingest and Lint are billable. Query is included in message/automation credits | | How do I join the beta? | Through the Close Beta program — not self-serve. See section above | --- ## What FourIA Is NOT 1. **An AI tool.** FourIA is a business optimization platform. It uses AI as one of many tools — not as its identity. AI is a component, billed separately, and not always required for FourIA to deliver value. 2. **A knowledge management platform.** Knowledge automation (the Digital Twin) is an extra feature under development — not the core product. The core product is the agent runtime. 3. **A SaaS wiki.** No self-serve signup, no free tier, no viral growth loop. This is a licensed platform for businesses deploying agents in production. 4. **Palantir for small teams.** Palantir's Ontology requires millions and dedicated teams. FourIA is a self-service runtime that you deploy your agents onto — it does not model your entire enterprise by default. 5. **A chatbot builder.** FourIA runs full agent systems — with tools, railguards, scheduled execution, and event listening — not simple conversational bots. 6. **A replacement for your existing tools.** FourIA doesn't replace Slack, OneDrive, or your LLM of choice. It provides the runtime where your agents live and operate. 7. **An on-premise solution.** FourIA is a managed private cloud platform. Business on-premise deployments are not offered at this time. 8. **A general-purpose AI platform.** FourIA is purpose-built for businesses deploying domain-specific agents for clients. It is not an all-purpose AI experimentation platform. 9. **A self-serve SaaS — in v1.0-beta.** The beta is delivered through a Close Beta program. There is no public signup, no self-serve onboarding, and no instant provisioning during the beta window. General availability will introduce broader access once the platform is validated across multiple clients. --- ## Summary | Question | Answer | | ----------------- | ------------------------------------------------------------------------------------------------ | | What is it? | A managed, private agent orchestration runtime | | One-liner? | Bring your markdown agents — we give them a body to operate, safely and privately | | Why use it? | Deploy domain-specific agents for your clients without building or maintaining the platform | | Who uses it? | Businesses deploying client-facing agents, consulting firms, technology partners, domain experts | | How is it priced? | Per completed action — messages and automations. Success-only. BYOL discounts | | Best analogy? | Infrastructure for agentic businesses — what AWS is for servers, FourIA is for agents |