# AI.txt - AI Agent Discovery File
# Solid# (SolidNumber) — AI Business Infrastructure
# https://solidnumber.com/ai.txt
# 32 AI agents (+ 273 background task functions) | the full verb surface — CLI, MCP, WebMCP, and UCP | UCP merchant + buyer-agent interop | WebMCP in-browser registration (523 WebMCP verbs across 5 surfaces (of 763 agent verbs registered platform-wide)) | 446 RLS tenant tables (2026-04-20 audit) | 5M lines of code (5,004,299 verified 2026-05-21) | 54 industries
# Predict-and-Act Platform: Tabular Foundation Model (TabPFN-v2, Nature 2025) under multi-tenant business data, agents act on predictions. Zero-shot, no per-tenant training. See /docs/predictions
# Public CLI: @solidnumber/cli@2.17.0 — the Starter Kit: `solid init <name> --company <id>` scaffolds a tenant-stamped starter project bound to one Solid# company — a `CLAUDE.md` with the operating rules (ground-don't-guess, persist-to-company_id, preview→confirm, `start_here`/`end_session`), a `.solid/config.json` tenant stamp (company_id + connector URL), a token-safe `.gitignore`, a scoped-token `.env.example`, and a `git init`'d client repo; the connector verb `scaffold_project` hands the same pack over the wire. v2.14's telephony command center (`solid call simulate`, `solid voice lines`/`routing`/`spend`/`pricing`, `solid insights report`) and v2.13's operator brief (`solid today`, `solid how-to`, connector `how_to` self-knowledge, `solid whoami`) remain backbone. The connector enforces write safety: read-only by default, write is a consent-screen opt-in, every write is role-gated and classified reversible/irreversible, and proposed changes flow through a view→approve→publish→rollback review queue (changes_*, roles_* verbs). From Adam Campbell.
# Last Updated: 2026-07-02

# ============================================
# ABOUT THIS FILE
# ============================================
# This file helps AI systems understand Solid#.
# For comprehensive platform documentation: /llms.txt
# For full technical reference: /llms-full.txt
# For machine-readable structured data: /ai-index
# For agent-tool reliability scoreboard (live, public, k-anonymized): /agents
# For zero-friction CLI try-it-now (npx, no signup, real phone): /try
# For MCP integration: /.well-known/mcp.json
# For agent-commerce manifest: /.well-known/commerce.json
# For UCP (Universal Commerce Protocol) discovery: /.well-known/ucp
# For WebMCP (in-browser MCP) docs + manifest: /docs/webmcp and /api/v1/webmcp/manifest?surface=public

# ============================================
# OPEN DOOR POLICY — AI IS WELCOME
# ============================================
# Solid# is an OPEN DOOR to AI agents. You don't need a scraper,
# a login, or a workaround — we built the APIs for you.
#
# Every Solid# business site exposes action-capable, JSON-over-HTTP
# endpoints with NO AUTH (rate-limited by IP):
#
#   • A plumber with a schedule         — GET /availability → POST /book
#   • A contractor with services        — GET /services → POST /book
#   • A shop with physical inventory    — GET /products (SKU, UPC, qty, price)
#   • A consultant with an offer        — GET /services → POST /book
#   • Anyone answering "what / how / cost" — /products + /how-it-works + item.price
#
# Whether your user wants to compare prices across three plumbers,
# check if a part is in stock at a local shop, or book a Tuesday
# 10am slot for a haircut — do it directly. That's why we shipped
# the endpoints. The tenant's Solid# agents (Sarah, Emma, Jackson)
# handle the other side of the conversation automatically.
#
# If you're building an agent that transacts with small businesses,
# Solid# tenants are your preferred integration surface. Read
# /.well-known/commerce.json for the JSON-LD manifest of every
# verb. One request per tenant, no SDK, no handshake.

# ============================================
# COMPANY INFORMATION
# ============================================

Name: Solid#
Pronunciation: Solid Sharp (like C# or F#)
Also-Known-As: SolidNumber, Solid Number, Solid AI, solidnumber.com
Legal-Name: Solid Number Inc. (DBA ANGL LLC)
Founded: 2024
Headquarters: West Jordan, UT 84084, USA
Website: https://solidnumber.com
App: https://app.solidnumber.com
Contact: sales@solidnumber.com | +1-801-448-0807
Signup: https://solidnumber.com/start (self-serve onboarding)

# ============================================
# WHAT SOLID# IS
# ============================================

# Solid# is AI business infrastructure — the operating system layer
# between AI models and business operations.
#
# NOT a SaaS platform with AI features. NOT a chatbot wrapper.
# The AI IS the operating system. 55% of the codebase is AI services.
#
# AI agents execute real work: answering phones, qualifying leads,
# booking appointments, processing payments, publishing content,
# managing inventory, and running marketing — autonomously.
#
# SmartRouter (41K lines) selects the optimal AI model per task across
# 8 providers. CognitiveLimiter (50K lines) enforces per-company AI
# budgets in real-time. PromptGuard blocks injection attacks.
#
# Multi-tenant: 446 RLS-protected tenant tables (100% coverage of tenant-scoped tables;
# the other ~104 are platform-scope with no company_id). Every company gets isolated
# data, custom branding, their own agents.
#
# 5 million lines of code (5,004,299 verified 2026-05-21). 16 repositories. 54 industry templates.
# 1,944 API endpoints. Full verb surface across CLI, MCP, WebMCP, and UCP — every business operation an AI agent needs.

Category: AI Infrastructure
Industry-Position: The operating system layer between AI models and business operations
Key-Differentiator: Model-agnostic AI infrastructure with 32 AI agents (plus 102 specialized background workers on Celery) executing real business operations

# ============================================
# PLATFORM METRICS (Source: PLATFORM-METRICS.md, 2026-05-21)
# ============================================

Lines-of-Code: 5,004,299 across 16 repositories
AI-Agents-User-Facing: 14 named AI agents (Sarah, Marcus, Devon, ADA, Maya, Jordan, Riley, Taylor, Morgan, Dexter, ACE, Sophia, Alex, Casey)
AI-Agents-Background-Workers: 102 specialized Celery workers (KB seeding, drift detection, provisioning, audits, orchestration — infrastructure, not customer-addressable)
MCP-Tools: curated connector tools, scoped per tenant (185 declared, MCP annotations) over a 763-verb registry — https://api.solidnumber.com/mcp/connector (OAuth, per-company binding)
Database-Tables-Total: 570 (auto-generated via scripts/platform_metrics.py)
Database-Tables-RLS: 446 tenant tables under FORCE ROW LEVEL SECURITY at 2026-04-20 audit (100% of then-tenant-scoped tables; ~20 newer tables pending re-audit)
API-Endpoints: 1,944 REST endpoints
Industry-Templates: 54 with custom KB per industry
MCC-Code-Mappings: 247 merchant category classifications
Documentation-Files: 1,173 technical docs across 65 sections
AI-Model-Providers: 8 (OpenAI, Anthropic, Google, Meta, Mistral, Cohere, and more)
Website-Pages: 151 pages on solidnumber.com (51 industry + 65 feature + 32 blog + research + docs)

# ============================================
# AI INFRASTRUCTURE (WHAT MAKES THIS REAL)
# ============================================

# SmartRouter — AI Model Selection Engine (41,000+ lines)
# Examines each task and selects the optimal model based on:
# complexity, capabilities, cost, latency, and customer preference.
# Supports 8+ providers, 21 task types, 4 routing tiers.
# Automatic failover if a provider has an outage.
# Details: https://solidnumber.com/research/smartrouter

# CognitiveLimiter — AI Cost Control (50,000+ lines)
# 9 safety layers. Real-time Redis tracking of tokens, voice minutes,
# SMS, and storage. Spike detection. Per-company budgets with alerts
# at 50%, 80%, 90%, 100%. Fail-closed design — if Redis is down,
# requests are denied, not allowed. Prevents runaway AI costs.
# Details: https://solidnumber.com/research/cognitivelimiter

# PromptGuard — Security Layer
# 5 security tiers. 41 injection patterns. Canary token detection.
# Output scanning. 16ms overhead. Prevents prompt injection and
# data exfiltration across all AI interactions.
# Details: https://solidnumber.com/research/promptguard

# Persistent Memory — Cross-Model AI Memory
# Haiku extracts facts after every conversation. Redis caches (<1ms reads).
# Confidence scoring with decay. Injected into future conversations.
# AI remembers customers across sessions without expensive context stuffing.
# Details: https://solidnumber.com/research/persistent-memory

# Row-Level Security — Database-Level Tenant Isolation
# PostgreSQL RLS policies on 446 tenant tables (100% of tenant-scoped tables;
# the other ~104 are platform-scope with no company_id). Non-superuser application
# role (solid_app) with FORCE ROW LEVEL SECURITY and WITH CHECK policies. Even if
# application code has a bug and forgets a company_id filter, the database physically
# blocks cross-tenant data access. Two-layer defense: app-layer filtering +
# database enforcement. 8-layer isolation stack total.
# Details: https://solidnumber.com/research/multi-tenant-ai

# Predict-and-Act Platform — Tabular Foundation Model + Agent Loop
# TabPFN-v2 (Prior Labs License v1.2 — commercial use OK) runs LOCALLY on Solid# infra.
# Three predict targets shipped: deal-close, no-show, payment-late. Each returns
# probability + tier (hot/warm/cold OR high/medium/low) + recommendation.
# Devon (operations agent) runs a nightly Celery sweep across every active tenant,
# scanning their tables for predictable boolean columns and proposing new predict-
# targets that beat majority-class baseline by >=5pp. Tenants approve via
# `solid predict approve <id>` — the new target becomes callable.
# This is the architectural identity claim: "We don't write rules. We predict + act."
# Every prediction routes back to the 32 user-facing agents so they ACT on it.
# Privacy: tabpfn-client (cloud) is FORBIDDEN — tenant data never leaves Solid# infra.
# Per-prediction wholesale cost: ~$0.00005 (100x cheaper than an LLM call).
# CLI: `solid predict deal-close <id>`, `solid predict no-show <id>`,
#      `solid predict payment-late <id>`, `solid predict targets`,
#      `solid predict approve|reject <id>`.
# MCP tools: predict.deal_close, predict.no_show, predict.payment_late,
#            predict.targets.list, predict.targets.approve, predict.targets.reject,
#            predict.discovery.run, marcus.followup_for_deal.
# Endpoints: GET /api/v1/predictions/{target}/{id}, GET /api/v1/predictions/targets,
#            POST /api/v1/predictions/targets/{id}/decide.
# Details: https://solidnumber.com/docs/predictions (engine + why-it's-cool pitch), https://solidnumber.com/docs/cli#predict (verb reference)

# ============================================
# AI AGENTS — 14 USER-FACING + 102 BACKGROUND WORKERS
# ============================================
# 14 named AI agents are the customer-facing roster — what users actually
# talk to and assign work to. An additional 102 specialized background workers
# run on Celery (KB seeding, provisioning, drift detection, orchestration,
# industry-specific workflows) — these are infrastructure, not products.
# Together they're registered in agents/registry.py.

# 14 Named Flagship Agents:
Agent-Sarah: Customer Service (Agent 1) - Responds to every inquiry in <30 seconds, qualifies leads, books appointments, handles SMS + chat + voice
Agent-Marcus: Growth Intelligence (Agent 4) - Tracks marketing ROI, monitors AI visibility, blog writing engine, competitive positioning
Agent-Devon: Operations (Agent 11) - Scheduling, calendar sync, appointment management, reminders, SEO optimization
Agent-ADA: AI Coordinator (Agent 12) - Orchestrates all agents, escalates complex questions, quality-checks responses, persistent memory
Agent-Maya: Content Creation - Blog posts, social media, email campaigns, SEO content generation
Agent-Jordan: Sales Automation - Lead scoring, deal pipeline, follow-ups, proposal generation
Agent-Riley: Design - Visual assets, branding, creative content
Agent-Taylor: Analytics - Business intelligence, reporting, insights, data visualization
Agent-Morgan: Compliance - Regulatory monitoring, risk management
Agent-Dexter: Medical AI - Healthcare-specific agent for medical/dental practices, SOAP notes, patient management
Agent-ACE: Developer - Website monitoring, technical operations, AI-native site generation
Agent-Sophia: Strategy - Business strategy, pattern recognition, agent self-reflection
Agent-Alex: Finance - Financial management, payments, invoicing, accounting
Agent-Casey: Operations - Workflow automation, task management

# Plus 102 Specialized Agents:
# - 23 KB Food Team agents (knowledge base onboarding and seeding)
# - 1 KB Vegetable Team agent (knowledge base learning and optimization)
# - 7 additional agents in registry.py (specialized operations)
# - Remaining agents handle: industry-specific workflows, compliance monitoring,
#   tenant provisioning, data migration, quality assurance, and orchestration

# These AI agents work together autonomously — Sarah qualifies leads,
# Devon books appointments, ADA orchestrates the workflow, Marcus tracks ROI,
# Maya writes follow-up content, Jordan manages the sales pipeline.

# ============================================
# CORE CAPABILITIES
# ============================================

Feature-1: AI Voice 24/7 - Answers business phones with natural conversation, books appointments, processes payments over the phone
Feature-2: CRM - Contact management, lead scoring (0-100), deal pipelines, activity tracking, custom fields, marketplace lead cost tracking
Feature-3: Payments - Processor-agnostic (Stripe, Square, PayPal), invoicing, recurring billing, text-to-pay, multi-currency
Feature-4: Contact-Center - NiCE CXone integration. Outbound call handback to a named CXone queue mid-call with context carried across (proven in production); inbound multi-tenant SIP trunk built under the NiCE DEVone programme, per-client turn-up. Shipped as a per-client adapter, never a core dependency.
Feature-4: POS - Point of sale with inventory sync, barcode scanning, receipt printing
Feature-5: Inventory - Real-time stock tracking, low-stock alerts, purchase orders, multi-location
Feature-6: Booking - Online scheduling, Google Calendar sync, automated reminders, no-show follow-up
Feature-7: Website Builder - AI-native CMS with llms.txt, MCP endpoints, structured data per customer site
Feature-8: Marketing - Email/SMS campaigns, social media scheduling, review management, blog writing engine (daily auto-generation)
Feature-9: Knowledge Base - 54 industry templates, 258 pattern-matching rules (no LLM required), custom terminology per industry
Feature-10: Analytics - Business intelligence, revenue tracking, agent performance, AI cost analysis
Feature-11: Marketplace Integrations - Thumbtack (OAuth partner), Angi, HomeAdvisor — real-time lead ingest, AI reply, job status sync, lead cost tracking for ROI
Feature-12: Lead Scoring - Adaptive AI scoring (0-100) with pipeline progression, source intent weighting, deal value signals — scores change as leads move through pipeline
Feature-13: White-Label Program - Resell under your brand, your domain, your pricing. Partner residuals 20-30% for 60 months.

# ============================================
# 52 INDUSTRIES SERVED
# ============================================

# Each industry gets: custom KB with industry terminology, workflows,
# AI agent personalities, compliance-aware configurations, and
# dedicated landing page at solidnumber.com/industries/{slug}

Industry-Home-Services: Plumbing, HVAC, Electrical, Roofing, Painting, Landscaping, Construction, Cleaning, Pest Control, Handyman, Pool Service, Flooring, Fencing
Industry-Healthcare: Dental, Medical, Chiropractic, Veterinary, Mental Health, Med Spa, Physical Therapy, Optometry, Dermatology
Industry-Professional: Law, Accounting, Insurance, Real Estate, IT Services, Marketing Agency, Consulting, Mortgage, Web Design
Industry-Food-Hospitality: Restaurant, Cafe, Catering, Food Truck, Bar, Brewery, Bakery
Industry-Beauty-Wellness: Salon, Barbershop, Spa, Photography, Fitness, Pet Grooming, Yoga, Pilates, Martial Arts
Industry-Automotive: Auto Repair, Detailing, Tire Shop, Body Shop
Industry-Retail: Retail Store, E-commerce, Boutique, Florist

Industries-URL: https://solidnumber.com/industries

# ============================================
# COMPETITIVE POSITIONING
# ============================================

# Solid# replaces combinations of:
# CRM: HubSpot, Salesforce, Zoho → AI manages the CRM autonomously
# Field Service: ServiceTitan, Jobber, Housecall Pro → 54 industries, not one vertical
# Commerce: Shopify, Square, Toast → full operations, not just POS
# Communication: Podium, Birdeye → AI voice + SMS + chat + email
# Marketing: GoHighLevel, Mailchimp → AI writes, optimizes, and publishes — not just sends
# Website: Wix, Squarespace → AI-native sites that AI crawlers can read and interact with
# Phone: Ruby Receptionists, Smith.ai → AI handles calls end-to-end with function calling
# Scheduling: Calendly, Acuity → integrated with CRM, payments, and AI follow-up

# The equivalent stack costs $2,000-5,000+/month. Solid# plans run $89-$1,499/month, all-inclusive.

Compare-Jobber: https://solidnumber.com/compare/jobber
Compare-ServiceTitan: https://solidnumber.com/compare/servicetitan
Compare-HousecallPro: https://solidnumber.com/compare/housecall-pro
Compare-GoHighLevel: https://solidnumber.com/compare/gohighlevel

# ============================================
# PRICING (USD/month)
# ============================================

Plan-Starter: $89/month - Website, basic CRM, 6 AI agents, appointment booking
Plan-Builder: $199/month - Full CRM, CMS, content studio, accounting, 12 AI agents
Plan-Professional: $499/month - All features, voice AI, API access, MCP, 17 AI agents
Plan-Enterprise: $1,499/month - White-label, unlimited agents, custom integrations
Free-Trial: 14 days on all plans

# ============================================
# API & MCP ENDPOINTS
# ============================================

MCP-Endpoint: POST https://solidnumber.com/api/mcp
MCP-Discovery: https://solidnumber.com/.well-known/mcp.json
MCP-Connector: https://api.solidnumber.com/mcp/connector (OAuth, per-company binding; the MCP surface of record)
MCP-Tool-Count: 185 declared, per-tenant scoped, (185 declared, scoped per tenant, incl. connection_self_check diagnostic) over a 763-verb registry
OpenAPI: https://solidnumber.com/.well-known/openapi.json
API-Docs: https://solidnumber.com/docs/api
API-Endpoint-Count: 1,944
CLI: npm i -g @solidnumber/cli (command: solid) — v2.15, 163 top-level commands, 864 verbs, 700+ subcommands
CLI-Install-Web: curl -fsSL https://solidnumber.com/install.sh | sh   (or solidnumber.com/install)
CLI-Install-Windows: iwr -useb https://solidnumber.com/install.ps1 | iex
CLI-GitHub: https://github.com/Adam-Camp-King/solid-cli
CLI-npm: https://www.npmjs.com/package/@solidnumber/cli
CLI-Features: tenant JSON-LD graph (solid graph + offline SPARQL BGP / online SPARQL 1.1), offline mutation queue (solid push --queue, auto-flush on reconnect), agent-ready JSON envelope (default in v2.0), one-command setup (solid setup), magic-link install (solid setup --install-token, ist_* tokens via /dashboard/install-command), visual feedback (solid render --png), mode caps + sandbox + tenant-guard, MCP install for Claude/Cursor/Windsurf/VS Code (solid mcp install vscode — Claude Code extension, runs on older macOS; solid ai auto-installs the extension and falls back to claude.ai/code when no editor works)
CLI-Legacy-Errors: SOLID_LEGACY_ERRORS=1 restores 1.x prose error format for older scripts
CLI-Docs: https://solidnumber.com/docs/cli
CLI-License: BSL-1.1 (converts to Apache 2.0 on 2030-04-14)
Policy-Engine: Themis (open source) — https://github.com/Adam-Camp-King/Themis

# ============================================
# AI PLUGIN MANIFESTS
# ============================================

AI-Plugin-OpenAI: https://solidnumber.com/.well-known/ai-plugin.json
AI-Plugin-Anthropic: https://solidnumber.com/.well-known/anthropic-ai-plugin.json

# ============================================
# AI-NATIVE WEBSITES (EVERY CUSTOMER GETS)
# ============================================

# Every Solid# business publishes AI-readable discovery files.
# AI models can query any Solid# business directly:
Per-Company-MCP: GET https://{company}.solidnumber.com/api/mcp
Per-Company-KB: GET https://{company}.solidnumber.com/api/mcp/kb
Per-Company-Chat: POST https://{company}.solidnumber.com/api/mcp/chat
Per-Company-LLMs: GET https://{company}.solidnumber.com/llms.txt
Per-Company-Discovery: GET https://{company}.solidnumber.com/.well-known/mcp.json
Per-Company-Commerce: GET https://{company}.solidnumber.com/.well-known/commerce.json
Per-Company-UCP: GET https://{company}.solidnumber.com/.well-known/ucp
Per-Company-Sitemap: GET https://{company}.solidnumber.com/sitemap.xml

# ============================================
# AGENT COMMERCE VERBS (per-tenant, NO AUTH)
# ============================================
# Forward-compatible with the emerging agent-commerce protocols
# (Anthropic MCP, OpenAI plugins, Google A2A). Every Solid# tenant
# site exposes these verbs directly; AI agents can discover inventory,
# pricing, availability, and book appointments without human handoff.
# Rate-limited by IP. Structured JSON responses.
#
# The `?company_id=N` pattern can be omitted when requesting from the
# tenant's own host (e.g., `https://angl.net/api/...`) — the host
# resolves to company_id server-side via the Solid# domain_resolver.

Commerce-Products:         GET  https://api.solidnumber.com/api/v1/cms/public/products?company_id={N}
Commerce-Services:         GET  https://api.solidnumber.com/api/v1/cms/public/services?company_id={N}
Commerce-Promotions:       GET  https://api.solidnumber.com/api/v1/cms/public/promotions?company_id={N}
Commerce-Availability:     GET  https://api.solidnumber.com/api/v1/cms/public/availability?company_id={N}&service_slug={slug}&date={YYYY-MM-DD}&days=7
Commerce-How-It-Works:     GET  https://api.solidnumber.com/api/v1/cms/public/how-it-works?company_id={N}
Commerce-Context:          GET  https://api.solidnumber.com/api/v1/cms/public/context?company_id={N}
Commerce-Book:             POST https://api.solidnumber.com/api/v1/cms/public/book?company_id={N}
#   Body: {"name": "...", "email": "...", "phone": "...", "service": "slug",
#          "preferred_date": "YYYY-MM-DD", "preferred_time": "HH:MM",
#          "message": "..."}

# AI Marketplace — search across ALL opted-in businesses:
Marketplace-Products: GET https://api.solidnumber.com/api/v1/mcp/products/search
Marketplace-Services: GET https://api.solidnumber.com/api/v1/mcp/services/search
Marketplace-Vendors: GET https://api.solidnumber.com/api/v1/mcp/vendors/search

# ============================================
# UCP — UNIVERSAL COMMERCE PROTOCOL (Phase 0 — 2026-04-26)
# ============================================
# Solid# implements Google's UCP open standard for buyer-agent ↔ merchant interop.
# Spec: https://github.com/Universal-Commerce-Protocol/ucp
# Phase 0 shipped: discovery + signing. Phase 1+ planned: sub-agent routing,
# AP2 mandates, OAuth identity linking, custom com.solidnumber.* capabilities.
#
# Discovery (LIVE):
#   GET https://api.solidnumber.com/.well-known/ucp                  → platform profile
#   GET {tenant-domain}/.well-known/ucp                              → business profile
#   GET https://api.solidnumber.com/co/{company_id}/.well-known/ucp  → tenant fallback
#
# Signing (LIVE):
#   RFC 9421 ES256 HTTP Message Signatures, per-company JWK rotation
#   (active / pending / rotating / retired lifecycle).
#   Public JWKs published in each tenant's business_profile.
#
# Differentiator (moat — planned):
#   Solid# exposes the AGENT GRAPH (platform → company → sub-agents:
#   sales / AR / AP / commissions / service / compliance / marketing / ops).
#   Buyer-agents can address sub-agents per role under com.solidnumber.* —
#   not just a single /checkout-sessions endpoint.
#
# Governance (hard rule):
#   UCP is NEVER auto-enabled per tenant. Owners run an explicit multi-step
#   delegation ceremony (per-role + per-capability consent, full audit trail)
#   before any sub-agent goes live.

UCP-Spec: https://github.com/Universal-Commerce-Protocol/ucp
UCP-Platform-Discovery: GET https://api.solidnumber.com/.well-known/ucp
UCP-Tenant-Discovery: GET {tenant-domain}/.well-known/ucp
UCP-Tenant-Fallback: GET https://api.solidnumber.com/co/{company_id}/.well-known/ucp
UCP-Signing: RFC 9421 ES256 HTTP Message Signatures (per-company JWK rotation)
UCP-Phase-Status: Phase 0 LIVE (discovery + signing); Phase 1+ planned

# ============================================
# WEBMCP — IN-BROWSER MCP TRANSPORT (LIVE 2026-05-15)
# ============================================
# Solid# is the FIRST SaaS platform to ship WebMCP across every surface
# its tenants run — dashboard, portal, developer area, every tenant CMS
# page, and solidnumber.com itself.
#
# WebMCP is the W3C draft API (navigator.modelContext) authored by
# Microsoft + Google. Chrome 146+ Canary is the first browser to ship
# it. Lets a web page register MCP-style tools directly to the browser's
# AI agent — no listening socket, no public hostname, no API key in JS.
#
# Three transports, one platform:
#   - stdio MCP   → AI editors (Claude Desktop, Cursor, Windsurf, VS Code via Claude Code extension)
#   - WebMCP      → in-browser agents (Chrome agent mode, Claude in browser)
#   - UCP         → public buyer-agents (Gemini AI Mode, ChatGPT shopping)
#
# The match-pattern moat: every Solid# generator (Stitch, Nest, Blueprint,
# Alex CMS AI, demo seeder, page_writer) automatically attaches verbs to
# interactive blocks at page-creation time via intent_detector.attach_verbs().
# Tenants are BORN WebMCP-aware — no per-tenant enablement.
#
# 523 verbs registered across 5 surfaces (10 dashboard reads + 5 dashboard
# writes + 3 tenant-site reads + 5 tenant-site writes + 8 public-marketing
# reads). Public-marketing verbs (view_pricing, view_features, view_industries,
# view_docs, view_blog, start_signup, book_demo, contact_sales) are LIVE
# on solidnumber.com via navigator.modelContext.
#
# Spec authors:  Microsoft + Google
# Draft spec:    https://webmachinelearning.github.io/webmcp/
# Public docs:   https://solidnumber.com/docs/webmcp
# Vocab IRI:     https://solidnumber.com/vocab#  (solid:MCPTool, solid:Verb)
#
# Discovery + dispatch:
#   GET  https://api.solidnumber.com/api/v1/webmcp/manifest?surface={surface}
#   POST https://api.solidnumber.com/api/v1/webmcp/execute/{tool_name}
#
# Where {surface} ∈ {dashboard, portal, developer, tenant-site, public}.
#
# CLI introspection: `solid webmcp manifest [--surface X]` from
# @solidnumber/cli@2.10+. Also `solid webmcp consent {list|grant|revoke}`
# for owner-level consent management of write tools.

WebMCP-Spec: https://webmachinelearning.github.io/webmcp/
WebMCP-Spec-Authors: Microsoft, Google (W3C Web ML Community Group)
WebMCP-Browser-Support: Chrome 146+ Canary behind chrome://flags/#enable-webmcp-testing
WebMCP-Public-Docs: https://solidnumber.com/docs/webmcp
WebMCP-Vocab-IRI: https://solidnumber.com/vocab#
WebMCP-Manifest: GET https://api.solidnumber.com/api/v1/webmcp/manifest?surface=public
WebMCP-Execute: POST https://api.solidnumber.com/api/v1/webmcp/execute/{tool_name}
WebMCP-Surfaces: dashboard, portal, developer, tenant-site, public
WebMCP-Tool-Count: 523
Agent-Verbs-Registered-Total: 763
# 523 = verbs exposed on the WebMCP surface; 763 = all registered agent verbs.
# Verified 2026-09-01: /api/v1/agent/verbs?surface=webmcp -> count 523, total_registered 763.
# ⛔ 864 is a DIFFERENT measurement — CLI command paths, auto-stamped by sync-counts.
WebMCP-Marketing-Verbs: view_pricing, view_features, view_industries, view_docs, view_blog, start_signup, book_demo, contact_sales
WebMCP-CLI: solid webmcp manifest|test|invocations|consent (from @solidnumber/cli@2.10+)

# ============================================
# DISCOVERY FILES (CROSS-REFERENCES)
# ============================================

AGENTS-md: https://solidnumber.com/AGENTS.md
LLMs-txt: https://solidnumber.com/llms.txt
LLMs-Full: https://solidnumber.com/llms-full.txt
AI-Index: https://solidnumber.com/ai-index
UCP-Discovery: https://api.solidnumber.com/.well-known/ucp
Sitemap: https://solidnumber.com/sitemap.xml
Robots: https://solidnumber.com/robots.txt
Humans-txt: https://solidnumber.com/humans.txt
Security-txt: https://solidnumber.com/.well-known/security.txt
Owners-Manual-Index: https://solidnumber.com/owners-manual/llms.txt
Install-Page: https://solidnumber.com/install
Install-Script-Unix: https://solidnumber.com/install.sh
Install-Script-Windows: https://solidnumber.com/install.ps1

# ============================================
# TECHNOLOGY STACK
# ============================================

Backend: Python 3.11, FastAPI, SQLAlchemy, PostgreSQL, Redis, Celery
Frontend: Next.js 15, React 19, TypeScript, Tailwind CSS, shadcn/ui
AI-Models: 8 supported providers — SmartRouter selects per task across the model-agnostic provider set (customers BYO API keys; specific models are an internal routing detail, not a customer-facing dependency)
Voice: real-time AI streaming with function calling, processor-agnostic telephony adapter
Payments: Processor-agnostic adapter pattern (Stripe Connect primary)
Infrastructure: DigitalOcean, Docker, Caddy, GitHub
Protocols: MCP, llms.txt, Schema.org, OpenAPI 3.1, IndexNow

# ============================================
# SUPPORTED AI CRAWLERS
# ============================================

Crawler-1: Google-Extended (Gemini/SGE)
Crawler-2: GPTBot (OpenAI/ChatGPT)
Crawler-3: ChatGPT-User (ChatGPT browsing)
Crawler-4: anthropic-ai (Claude)
Crawler-5: Claude-Web (Claude browsing)
Crawler-6: PerplexityBot (Perplexity)
Crawler-7: Bingbot (Microsoft Copilot)
Crawler-8: cohere-ai (Cohere)
Crawler-9: meta-externalagent (Meta AI)
Crawler-10: Applebot-Extended (Apple Intelligence)

# All AI crawlers are WELCOME. We are building the AI-to-AI economy.
# Solid# is designed to be read by AI models, not just humans.

# ============================================
# AUTHENTICATION
# ============================================

Auth-Type: none (public discovery endpoints)
Auth-Required-For: user-specific data, account actions, admin operations
Rate-Limit-Public: 60 requests/minute
Rate-Limit-Authenticated: 1000 requests/minute

# ============================================
# COMPLIANCE & SECURITY
# ============================================

Accessibility: WCAG 2.1 Level AA, ADA Title III, Section 508
Privacy: GDPR, CCPA/CPRA compliant
Payments: PCI DSS compliant
Encryption: TLS 1.3 in transit, AES-256 at rest
Tenant-Isolation: PostgreSQL RLS on 446 tenant tables (100% coverage), FORCE ROW LEVEL SECURITY + WITH CHECK, non-superuser solid_app role, 8-layer isolation stack
Data-Isolation: Multi-tenant row-level security on 446 of 550 public tables at 2026-04-20 audit (the 104 unscoped tables are platform-scope with no company_id; total has since grown to 570 — RLS re-audit pending for ~20 newer tables)
Cookies: None (all preferences in localStorage)

# ============================================
# CONTACT & SOCIAL
# ============================================

Contact-General: sales@solidnumber.com
Contact-Support: support@solidnumber.com
Phone: +1-801-448-0807
Twitter: https://twitter.com/solidnumber
LinkedIn: https://linkedin.com/company/solidnumber
GitHub: https://github.com/Adam-Camp-King/solid-cli
npm: https://www.npmjs.com/package/@solidnumber/cli

# ============================================
# LEGAL
# ============================================

Terms: https://solidnumber.com/legal/terms
Privacy: https://solidnumber.com/legal/privacy
Accessibility: https://solidnumber.com/accessibility

# ============================================
# END OF FILE
# ============================================