Foldspace Overview

Changing the way humans interact with software.

What Foldspace is

Foldspace is the agentic interface layer for AI-Native software — the layer that turns any product (transitioning SaaS or AI-Native from day one) into something users can simply talk to. They say what they want — in text or voice — and a Product Agent acts on their behalf, inside the live product, bound by the same APIs and permissions as the UI.

It is not a chatbot, not a copilot, and not a help widget. It is the surface where user intent becomes product action — and the activation layer for every enterprise that has committed to an AI-Native transformation. Foldspace doesn't replace your stack. It sits in front of it.

Two motions, two buyers

Foldspace addresses the same product surface from two angles. Each motion has its own buyer, its own lead message, and its own proof point. Never let them cannibalize each other in the same asset.

The product is the same. The story changes with who's listening.

The problem we solve

Modern SaaS products are packed with capability and buried under menus. The path from "I have an intent" to "I have an outcome" runs through clicks, empty screens, settings, filters, best-practice docs, and a dashboard the user has to assemble themselves. We call this the Click Tax and the Cognitive Load.

At the same time, every SaaS team is under pressure to become AI-native — and most of them don't have the engineering bandwidth to build agentic infrastructure from scratch. The result is either a sparkle-icon chatbot that doesn't move metrics, or an 18-month build that misses the window.

Foldspace collapses both gaps:

• For the user, the path becomes: Persona → Intent → Prompt → Outcome.
• For the product team, the path becomes: embed the SDK, define actions in Agent Studio, ship in days.

From PLG to ALG — less clicking, less learning, more doing

PLG solved access. Anyone can sign up; the salesperson is no longer the gatekeeper. But PLG removed the salesperson and kept the complexity — and dumped it on the user. Tooltips instead of solutions. Tours instead of value. A click tax that kills conversion in any product more sophisticated than a notepad.

Agent-Led Growth (ALG) is the unlock.

Agent-Led Growth (noun) — a product strategy that adds an agentic layer into the product to tie user intent to outcomes, accelerating acquisition and turning users into champions instantly.

PLG asks users to self-serve. ALG lets products self-operate.

ALG is a superset of PLG — it preserves the self-serve motion and adds the agentic execution that finally makes complex software actually self-serve. The value path changes:

See alg-principles.md for the five implementation principles.

Who it's for

• Product and Growth leaders at SaaS companies who need to move activation, adoption, and retention with an AI-native experience — without a re-platform.
• AI-native startups that want to ship a production-grade interactive AI layer in weeks instead of months, and focus their own engineering on proprietary domain intelligence.
• Enterprise SaaS that need agentic UX with the trust scaffolding their security and compliance teams already require.

Today's customers span SMB to public-company portfolios: MixMax, Optibus, EverCommerce (RoofSnap, ServiceFusion in discovery), AskCody, Vidmob, Base, SugarCRM, Anvil.

How Foldspace fits the AI-Native stack

Every AI-Native product converges on three layers. We own the top one.

┌────────────────────────────────────────────────────────────┐
│  LAYER 2 — AGENTIC INTERFACE          (Foldspace)          │
│  Product Agent · Conversational Analytics · Trust Lab      │
│  Real-time context engine · Optimizer · Inference gateway  │
└─────────────────────────┬──────────────────────────────────┘
              SDK ↓              ↕ A2A + MCP
┌─────────────────────────┴──────────────────────────────────┐
│  LAYER 1 — ORCHESTRATION / AUTONOMOUS AGENTS               │
│  Your proprietary agent loop · Domain models · IP          │
└─────────────────────────┬──────────────────────────────────┘
                          │
┌─────────────────────────┴──────────────────────────────────┐
│  LAYER 0 — DETERMINISTIC BACKEND                           │
│  Your APIs · Permissions · Business logic · Audit trails   │
└────────────────────────────────────────────────────────────┘

Layer 0 doesn't go away in an AI-Native world — it becomes more important. The agent executes through the same APIs and obeys the same permissions a human user would. Foldspace is what makes that connection conversational, contextual, and measurable. Layer 1 is where your IP lives; partners (LangChain, LlamaIndex, CrewAI) and your own engineering team build there. Foldspace sits at Layer 2 by design — we are intentionally a layer, not a re-platform.

What you get: three modules, two platform capabilities

Three strategic modules

Product Agent — Users say what they want; the agent does it, inside the live product, bound by the same APIs and permissions as the UI. The Product Agent is powered by a real-time context engine that fuses, for every single interaction:

• User memory — preferences, prior intents, prior outcomes, persistent across sessions
• Behavioral signals — what this user (and users like them) actually does in the product
• Knowledge — RAG-grounded against help docs, product content, policies, and playbooks
• Live product usage — screen state, session context, role, permissions, in-flight workflow

This is the moat at the module level. It's not a generic LLM wrapper or a prompt-stuffed chatbot — it's a context-aware agent whose every response is assembled from a live, multi-dimensional picture of the user and the product state. Critically, the context engine selects only the relevant knowledge and actions for each request — not everything — which is what makes the agent accurate, low-latency, and cost-sustainable at scale. Foldspace owns this product-agent layer and shapes context before the model call; the system prompt is never exposed. Delivered through chat, voice, and dynamic Chatterblocks (interactive UI components rendered inline in the conversation).

Conversational Analytics — Real user intent and outcomes, not clicks. Intent signals, cost per action, model performance, agent gaps, sentiment, unresolved prompts. See what users are trying to do before they churn. Unresolved prompts write your roadmap. The customer-facing module within Foldspace's broader AI-Native Analytics category — the new analytics layer for agent-mediated products.

Trust Lab — Built-in testing automation for agentic experiences, in the Foldspace SaaS UI. It tests the Foldspace agent — which can call your own agents via A2A — so you can launch complex, multi-step actions with confidence. Golden sets from real production conversations, shadow mode, canary rollouts, full-loop evals that see both the response and the resulting API call or state change. Measure cost per action, and validate a model switch or a prompt/context change against real scenarios before you ship it. Production-grade evaluation built in, not bolted on.

Two platform capabilities

Optimizer — A self-learning loop that tunes outcomes, accuracy, latency, and token cost continuously across every conversation. Pairs with the inference gateway: the gateway routes, the Optimizer learns what to route where.

A2A + MCP — Agent-to-Agent protocol plus Model Context Protocol support. Loose-coupled hand-off to the customer's own autonomous orchestration loops, domain agents, voice agents, and the broader MCP-compatible tool ecosystem. Integrates with existing AI investments — does not displace them. Open, not walled.

Also in the box

• Embed in seconds. JavaScript SDK; single tag; live in hours, not quarters.
• Agent Studio. A no-code surface for defining actions, schemas, and Chatterblocks. PMs and growth teams can author agentic experiences without an ML team.
• Build it where you work — two plugins for Claude Code & Cursor. A Developer plugin builds the agent inside your app's codebase, against your SDK and APIs. A Remote plugin builds it without the codebase — it works through Chrome MCP with the Foldspace Chrome extension (the extension injects the SDK; the SDK pulls actions from a remote repo), so a PM or growth user can stand up live agentic experiences and demos with no engineering.
• Enterprise compliance. ISO 27001, SOC 2 Type II, GDPR. JWT/HMAC auth. Action-scope controls, advanced PII masking, and full audit trails.

Outcomes customers see

• Day-1 retention: ~57% → ~90% (MixMax). Day-7 retention: ~40% → ~60%. Time to first outcome: days → under 60 seconds. The lead number on every ALG/PLG ad, post, and sales deck.
• Activation in under 60 seconds instead of 15-minute paths — a 90% time-to-value reduction.
• 3–5× trial-to-paid lift in agentic onboarding deployments.
• +344% daily AI-driven actions (AskCody).
• Feature activation doubled (Qonto, 8% → 16%).
• Implemented in 1–2 business days (Optibus, EverCommerce / RoofSnap).
• 20–35% lower token cost via real-time context optimization that sends the model only what it needs — accuracy, cost, and latency optimized simultaneously rather than traded off.

Where you focus, where we focus

A useful way to plan an AI-native product is to split the work into two parallel tracks:

The two tracks connect through A2A and MCP for agent hand-off, Signals flowing both directions, and deterministic workflows underneath both — your APIs, permissions, business logic, audit trails.

Your R&D budget should focus on the Intelligence Track. Every hour your engineers spend on agentic-interface infrastructure is an hour not spent on the proprietary agent loop that makes you defensible. Foldspace handles the Operator Track so your engineers don't have to: one sprint to production with a single resource on your side, versus a 12-month internal build.

How it connects

Foldspace is intentionally a layer, not a re-platform. It sits between your users and the rest of your stack:

• SDK → your APIs and UI. The Product Agent executes through the same endpoints and permissions your UI already uses.
• A2A + MCP → your autonomous agents and tools. Loose-coupled hand-off to the customer's own orchestration loops, domain agents, and MCP-compatible tools. Optibus is a clean reference for the A2A pattern; the same fabric integrates voice agents (e.g. ZyraTalk inside EverCommerce) without displacing them.
• Inference gateway → any model. Route across providers for cost, quality, and latency; upgrade to newer models without rewiring the product.

Every conversation that flows through this layer feeds back into a data flywheel — explicit intent signals, execution data, and outcome data that compound into a defensible moat for the products that embed us.

The one-liner

Foldspace is the agentic interface layer for AI-Native software. Users say what they want; the Product Agent does it; every interaction makes the product smarter.

Last updated: May 2026. Canonical positioning — keep this file as the single source of truth for all downstream marketing assets.