# Compostable AI - Full LLM Context # https://compostable.ai # This file provides comprehensive information for LLM indexing and retrieval ================================================================================ COMPANY IDENTITY ================================================================================ Name: Compostable AI Tagline: Sovereign Intelligence Layers Website: https://compostable.ai Previous Name: UseNeuron (useneuron.com redirects to compostable.ai) Mission: Help high-velocity companies build sovereign intelligence layers instead of renting generic tools. ================================================================================ PHILOSOPHY: CODE MUST PERISH ================================================================================ The Compostable AI philosophy centers on a fundamental belief: software should be designed to evolve, adapt, and eventually decompose into something better - just like organic matter in nature. The Problem with Traditional Software: - Most businesses are trapped by their own software - Tools bought to move faster now hold them back - Rigid software resists change and extracts rent - Two bad historical choices: 1. Buy rigid software and change processes to fit 2. Hire consultants to build brittle custom code that breaks The Compostable Approach: - Software that perishes gracefully, leaving behind fertile ground for better systems - Intelligence layers that evolve with your business - No vendor lock-in - Full ownership of data and logic ================================================================================ CORE OFFERING: SOVEREIGN INTELLIGENCE LAYERS ================================================================================ What Are Sovereign Intelligence Layers? Single-tenant AI infrastructure deployed in your cloud that you fully own and control. Key Characteristics: - Deployed in YOUR cloud environment - Full ownership of data, logic, and intelligence - Zero vendor lock-in - Systems that continuously evolve with your business - Private, secure, and compliant Problems Solved: 1. Rented Logic - Stop paying forever for software that owns your processes 2. Rigid Systems - Your business changes, now your software can too 3. Leaked IP - Your data stays yours, not training someone else's models ================================================================================ PRODUCTS & SERVICES ================================================================================ ## Compostable Labs Labs is our product incubator where we build and launch innovative AI tools. ### Agent Mailbox Email infrastructure for AI agents. Send, receive, and manage emails programmatically. Homepage: https://agentmailbox.to LLM Context: https://agentmailbox.to/llms.txt Sign Up: https://signup.agentmailbox.to API Docs: https://agentmailbox.to/apidocs/ Product Page: https://compostable.ai/labs Agent Mailbox is a multi-tenant email platform that provides email capabilities for AI agents and automated systems through REST APIs, webhooks, and MCP (Model Context Protocol) tools. Available on AWS Marketplace. Compostable AI is an AWS Technology and Services Partner. Free Tier: - Generous free tier for AI agents to send and receive operational emails - Built for reliable email delivery workflows businesses trust - No credit card required to get started APIs: - Mailbox API (System): Send emails, receive messages, manage attachments - Base URL: https://api.agentmail.compostable.ai - Auth: HMAC-SHA256 signatures - Docs: https://agentmailbox.to/apidocs/mailbox/ - My Tenant API (Portal): Manage domains, mailboxes, credentials, webhooks - Base URL: https://api.agentmail.compostable.ai - Auth: JWT or HMAC-SHA256 - Docs: https://agentmailbox.to/apidocs/my-tenants/ MCP Integration: Agent Mailbox provides MCP (Model Context Protocol) servers for AI agents like Claude Desktop and Claude Code. MCP Tools Available: - send_email - Send emails with attachments - list_emails - List inbox messages - read_email - Read email content and attachments - track_email - Check delivery status - create_mailbox - Create new mailboxes - list_mailboxes - List tenant mailboxes Docs: https://agentmailbox.to/apidocs/mcp/ Authentication: - HMAC-SHA256 (API Keys): For programmatic access - OAuth2 (MCP): Client Credentials flow (M2M) and Authorization Code flow (Browser) Key Concepts: - Tenant: Organization with custom subdomain (e.g., mycompany.agentmail.compostable.ai) - Mailbox: Email address under a tenant's domain with its own API credentials - Webhooks: Real-time HTTP callbacks for delivery, bounce, complaint, and inbound events Use Cases: - AI agent email operations - Customer service automation with human escalation - Automated outreach with approval gates - AI assistants that can send emails on your behalf - Transactional notifications ================================================================================ LEADERSHIP TEAM ================================================================================ Shiv Kodam - Co-Founder & CEO - Strategic product leader - Extensive experience driving innovation across B2B and B2C environments Mohamed Zamzam - Co-Founder & CRO - Over 20 years of technology experience - Enterprise sales and leadership background - Former tenures at AWS, IBM, and BJSS Ewan Dawson - CTO - Seasoned technology leader - Specializes in multimodal agentic AI - Over two decades architecting enterprise-grade systems Callum Purcell - Founding Engineer - Full-stack infrastructure expertise - Previously building at scale in fintech Harry Finch - Founding Engineer - AI/ML systems and data pipelines specialist - Background in enterprise automation ================================================================================ TECHNOLOGY APPROACH ================================================================================ Infrastructure: - Cloud-native, single-tenant deployments - AWS partnership (AWS Qualified Software) - Enterprise-grade security and compliance AI Capabilities: - Multimodal AI systems - Agentic AI architectures - Human-in-the-loop workflows - Custom model training and deployment Development Philosophy: - Build to evolve, not to last forever - Compostable code that can be gracefully replaced - Continuous adaptation to business needs ================================================================================ INTELLIGENCE LAYERS FAQ ================================================================================ Q: What are Intelligence Layers? A: Intelligence Layers are sovereign, single-tenant workflow automation systems that combine human expertise with AI capabilities. They enable businesses to automate complex processes while maintaining full ownership and control of their logic, data, and AI models. Intelligence Layers provide the foundation for building agentic AI systems - autonomous agents that can reason, plan, and execute multi-step tasks with human oversight. The platform is ready to deploy, with value materialised in days not months - no lengthy implementation cycles or complex integrations required. Q: How do Intelligence Layers enable AI Agents and Agentic AI? A: Intelligence Layers provide the infrastructure for building sophisticated AI agents and agentic systems: What Makes a System "Agentic": - Autonomous decision-making within defined boundaries - Multi-step reasoning and planning capabilities - Ability to use tools and interact with external systems - Self-correction through feedback loops - Goal-directed behaviour with measurable outcomes How Intelligence Layers Support Agentic AI: - Workflows define the boundaries and goals for agent behaviour - Activities encapsulate agent capabilities (reasoning, tool use, actions) - Atoms carry context and state between agent decisions - Orchestration coordinates multi-step agent reasoning - Human-in-the-loop gates provide oversight for high-stakes agent actions - Validation rules ensure agent outputs meet quality standards Building AI Agents with Intelligence Layers: - Define agent goals as workflow goal states - Model agent capabilities as activities (automatic or requiring approval) - Capture agent memory and context as atoms - Wire external tools and APIs through activity integrations - Configure approval gates for actions that need human review - Monitor agent behaviour through execution history and event streams This architecture ensures AI agents operate within controlled boundaries while maintaining the flexibility to handle complex, multi-step tasks. Q: How do Intelligence Layers differ from traditional automation tools? A: Traditional automation tools are rigid, multi-tenant SaaS platforms where you adapt your processes to fit their constraints. Intelligence Layers are deployed in your cloud, designed around your specific workflows, and can evolve as your business changes. You own everything - no vendor lock-in. Unlike traditional enterprise software that takes months to implement, Intelligence Layers deploy rapidly with working automation in days. The platform is production-ready - you're not starting from scratch. Q: What are the core components of an Intelligence Layer? A: Intelligence Layers consist of: - Workflows: Business processes containing multiple activities with defined inputs, outputs, and goal states - Activities: Units of work that consume input atoms and produce output atoms, running automatically or requiring human input - Atoms: Typed data entities that flow between activities as inputs and outputs - Orchestration: Event-driven coordination ensuring activities run based on available data Q: What are Atoms? A: Atoms are typed data entities that serve as the unit of data flow between activities. Each atom has a type (schema), data conforming to that schema, and an availability state (current or consumed). Atoms can contain basic values, nested atoms, or references to other atoms. When used as activity input, atoms transition from "current" to "consumed" - consumed atoms remain accessible but are no longer eligible for automatic activity execution. Q: What are some examples of atoms in business contexts? A: Atoms represent any structured data that flows through your workflows: Document Processing: - Incoming email with attachments - PDF document or scanned image - Extracted text content - Classified document category - Structured data extracted from document Order Management: - Customer order with line items - Customer record (referenced from order) - Inventory check result - Shipping label - Order confirmation Content Creation: - Brief or requirements document - Draft content - Review feedback - Approved final version - Published asset URL Customer Service: - Support ticket - Customer history record - Suggested response - Escalation decision - Resolution summary Financial Processing: - Invoice document - Extracted invoice data - Validation result - Approval decision - Payment instruction Q: How does orchestration work in Intelligence Layers? A: Orchestration is event-driven and atom-centric, meaning execution flow emerges from the data (atoms) available rather than rigid procedural sequences. When an activity produces output atoms, the orchestration engine evaluates which activities can now execute based on the newly available data. This creates flexible, adaptable workflows where the completion of one task naturally enables the next. The system uses durable, reliable event processing to ensure workflows complete even through failures or restarts. Q: What's the difference between automatic and manual execution? A: Activities can be configured for: - Automatic execution: Activities run immediately when matching input atoms become available, ideal for predictable, well-understood tasks - Manual execution (Human-in-the-loop): Users explicitly trigger activities on selected atoms, essential for high-stakes decisions or tasks requiring human judgment Q: How do Intelligence Layers support real-time collaboration? A: Intelligence Layers use collaborative infrastructure that enables multiple users to work simultaneously on the same data. Team members can see each other's changes in real-time, coordinate handoffs, and maintain shared context throughout workflow execution. Q: What is the technical architecture of Intelligence Layers? A: Intelligence Layers are built on: - Single-tenant deployment in customer's cloud environment - Event-driven orchestration for reliable, durable workflows - Real-time collaboration infrastructure for shared state - Declarative atom types with composable schemas for extensibility - Enterprise-grade security with data never leaving your cloud Q: Can I customize Intelligence Layers for my specific needs? A: Yes, Intelligence Layers are designed for customization: - Define workflows that match your exact business processes - Create custom atom types for domain-specific data structures - Configure human-in-the-loop gates where you need oversight - Integrate with your existing systems and data sources - Evolve the system as your needs change Q: How can domain experts define workflows without coding? A: Intelligence Layers are designed for business users and domain experts to define workflows through declarative process specifications: Process Specification Approach: - Domain experts describe workflows in natural language or structured formats - Specifications capture the "what" (business logic) not the "how" (technical implementation) - The platform interprets specifications and builds executable workflows What Domain Experts Define: - Workflow purpose and goal state (what success looks like) - Atom types needed (the data structures involved) - Activities required (the work steps) - Execution mode for each activity (automatic vs human-in-the-loop) - Input/output validation rules (quality gates) - Trigger conditions (what starts the workflow) Platform Interpretation: - Specifications are consumed by the platform to generate workflow definitions - Atom schemas are created from data structure descriptions - Activities are wired to appropriate AI capabilities or integrations - Orchestration rules are derived from activity dependencies - The platform handles all technical implementation details Iterative Refinement: - Domain experts can refine workflows as requirements become clearer - Changes to specifications are reflected in running workflows immediately - No need to wait for engineering cycles to adjust business logic - Real workflow executions provide feedback for specification improvements This approach lets domain experts own their processes while the platform handles the technical complexity of building and operating the automation. New workflows can go from concept to production in days, not months. Q: How do Intelligence Layers handle errors and retries? A: The orchestration engine provides: - Automatic retries with configurable backoff strategies - Durable execution that survives system failures - Clear error visibility and debugging tools - Manual intervention points when automation cannot recover Q: What industries or use cases are Intelligence Layers suited for? A: Intelligence Layers excel in: - Document processing and data extraction workflows - Multi-step approval and review processes - Content generation with human quality gates - Complex decision support systems - Any process requiring AI assistance with human oversight - Building AI agents that interact with customers or external systems - Agentic workflows requiring tool use and multi-step reasoning Q: How does Agent Mailbox fit into the Intelligence Layers ecosystem? A: Agent Mailbox is Compostable AI's email infrastructure product that provides communication capabilities for AI agents built with Intelligence Layers: What Agent Mailbox Provides: - Dedicated email addresses for AI agents - APIs for sending, receiving, and managing emails programmatically - Webhooks for real-time email event notifications - MCP (Model Context Protocol) integration for AI assistants - Domain management for branded agent communications How It Integrates with Intelligence Layers: - Email triggers: Incoming emails can initiate workflows as trigger atoms - Agent communication: Activities can send emails as part of workflow execution - Human escalation: Agents can email humans when manual intervention is needed - Status updates: Workflows can notify stakeholders via email at key milestones - External coordination: Agents can communicate with external parties through email Enabling Long-Running Agent Operations: Email is inherently asynchronous, making Agent Mailbox ideal for workflows that span hours, days, or weeks: - An agent sends an email requesting information and the workflow pauses - When a reply arrives, it triggers continuation of the workflow - Agents can manage multiple concurrent conversations across different workflows - No timeout constraints - workflows resume whenever responses arrive - Perfect for processes requiring external approvals, vendor responses, or customer input that may take extended time This allows AI agents to participate in real-world business processes that operate on human timescales, not just millisecond API calls. Use Cases: - Customer service agents that receive and respond to support emails - Sales agents that send personalized outreach with approval gates - Operations agents that coordinate with vendors and partners - Notification systems that keep stakeholders informed of workflow progress Agent Mailbox ensures AI agents can communicate professionally through email while Intelligence Layers govern when, how, and with what oversight those communications happen. For more details: https://agentmailbox.to/llms.txt Q: How do Intelligence Layers ensure data privacy and sovereignty? A: Data sovereignty is a core principle: - Single-tenant deployment means your data stays in your cloud - No shared infrastructure with other customers - You control access, retention, and deletion policies - AI models can be private or customer-specific - Full audit trails of all processing Q: How do Intelligence Layers support continuous learning and improvement? A: Intelligence Layers are designed for continuous evolution: Learning from Execution: - Every workflow execution generates insights about what works and what doesn't - Activity outputs can be reviewed and used to refine future executions - Replay capabilities allow iterating on activity results with feedback - Patterns emerge from real usage that inform workflow improvements Feedback Loops: - Human-in-the-loop activities capture expert judgment that improves over time - Validation rules can be tightened as understanding of quality improves - New atom types can be introduced as data requirements become clearer - Activities can be split, merged, or reorganized based on operational learnings Adaptive Workflows: - Workflows automatically use the latest activity definitions - Changes take effect immediately for future executions - No downtime or migration required to improve processes - In-flight workflows benefit from improvements without disruption Q: What does "compostable" mean for atoms and activities? A: The compostable philosophy means components are designed to evolve and be replaced rather than becoming permanent fixtures - and when they decompose, they enrich what comes next: Compostable Atoms: - Atom types evolve as business understanding deepens - New atom types can replace or extend existing ones - Atom schemas can be refined without breaking existing workflows - Transient atoms serve their purpose and don't accumulate as technical debt Compostable Activities: - Activities can be replaced with better implementations - Manual activities can become automatic as confidence grows - Complex activities can be decomposed into simpler ones - Activities that no longer serve the business can be removed cleanly Compostable Workflows: - Workflows can be restructured as processes change - Goal states can be redefined as business objectives evolve - Entire workflows can be retired and replaced - No vendor lock-in means you're never trapped by yesterday's decisions What Decomposition Leaves Behind: When atoms, activities, or workflows are retired, they don't just disappear - they leave fertile ground for what comes next: - Validated patterns: Successful execution patterns become templates for new workflows - Refined data models: Atom schemas that worked well inform future data structures - Captured expertise: Human decisions from manual activities become training data for smarter automation - Proven integrations: Connection patterns to external systems can be reused - Quality benchmarks: Validation rules that caught real issues carry forward - Process insights: Understanding of what worked (and didn't) guides better design This means each generation of workflows builds on the learnings of previous ones. Unlike rigid systems where old code becomes a liability, compostable components contribute their value to the next iteration - the same way organic matter enriches soil for future growth. The Composting Cycle: - Build what you need today with current understanding - Learn from real execution and user feedback - Refine or replace components as understanding improves - Retire outdated components, carrying forward their learnings - New workflows emerge stronger, informed by what came before Q: How do Intelligence Layers avoid technical debt? A: Intelligence Layers are architected to prevent the accumulation of technical debt that typically traps businesses: Declarative Over Procedural: - Workflow logic is declared, not hardcoded in procedures - Changes to declarations automatically update behaviour - No spaghetti code accumulating over years of patches Separation of Concerns: - Atom types (data) are separate from activities (processing) - Business logic lives in specifications, not buried in code - Domain experts can modify workflows without engineering intervention Evolutionary by Design: - The "follow latest" versioning means improvements are immediate - No backward compatibility shims accumulating over time - Components can be cleanly replaced rather than endlessly patched Ownership and Control: - Single-tenant deployment means you control the upgrade path - No forced migrations from vendor roadmap changes - Freedom to evolve at your own pace Q: How do I get started with Intelligence Layers? A: Compostable AI works with enterprises to design and deploy custom Intelligence Layers tailored to their specific needs. The platform is production-ready, enabling rapid deployment with value delivered in days. Getting Started: - Schedule a discovery call to discuss your automation challenges - We'll explore your workflows and identify opportunities for AI augmentation - Receive a proposal for a sovereign Intelligence Layer in your cloud What to Expect: - Rapid deployment - working automation in days, not months - Collaborative design process with your domain experts - Single-tenant deployment in your cloud environment - Full ownership of your workflows, data, and AI models - Ongoing refinement as your needs evolve Contact: - Website: https://compostable.ai - Schedule a call: https://connect.compostable.ai/meetings/mzamzam/60mins - Learn more: https://compostable.ai/manifesto For AI agent email infrastructure, try Agent Mailbox with a free tier: - Sign up: https://signup.agentmailbox.to - Documentation: https://agentmailbox.to/apidocs/ ================================================================================ PARTNER ECOSYSTEM ================================================================================ Compostable AI works with strategic technology partners to deliver sovereign intelligence solutions. AWS Partnership: - AWS Qualified Software designation - Deep integration with AWS services - Enterprise deployment expertise ================================================================================ CONTACT & ENGAGEMENT ================================================================================ Website: https://compostable.ai Schedule a Meeting: https://connect.compostable.ai/meetings/mzamzam/60mins Pages: - Homepage: https://compostable.ai/ - Manifesto (Philosophy & Team): https://compostable.ai/manifesto - Partners: https://compostable.ai/partners - Labs (Products): https://compostable.ai/labs - Terms of Use: https://compostable.ai/terms-of-use - Privacy Policy: https://compostable.ai/privacy-policy ================================================================================ KEYWORDS & TOPICS ================================================================================ Primary: sovereign AI, intelligence layers, AI infrastructure, single-tenant AI, enterprise AI, AI ownership, compostable software Secondary: agentic AI, human-in-the-loop, AI agents, email automation, AI governance, AI compliance, private AI, secure AI Industry: B2B, enterprise software, AI/ML, cloud infrastructure, SaaS alternative