Technical

Architecture

Sophona is an agentic automation platform built around a Skill Map graph. The graph dynamically gates context so the agent has access only to skills relevant to the current business path — instead of a flat prompt stuffed with every tool. Each skill is a self-contained workflow runtime that can call AI in real time (planning, tool routing, retries) and apply Mind Inject procedures at the right moment, giving you full control over how the agent behaves without hardcoded logic.

One mind, many channels

The same agent operates through phone, embedded web widget, desktop widget, and API/iPaaS — one runtime, multiple interfaces.

Skill Map is a graph (context gating)

Instead of a flat list of tools, Sophona uses a skill graph. At any moment the agent sees only skills relevant to the current business path — reducing context and improving reliability.

Each skill is its own workflow runtime

Every skill is an executable workflow that can call AI in real time (planning, tool routing, retries) and apply Mind Inject inside the skill — not only at the top-level agent.

Mind Inject (full business control)

You can inject procedures and policies at runtime — globally, per business path, or per skill. This enables deterministic, auditable behavior without hardcoding logic.

Autonomous actions (web + desktop)

Skills can execute real work: clicking, navigation, form filling, desktop operations, and API calls — verified and repeatable.

Pro instances: isolated single-tenant

Pro (Founders/Business) instances run in isolated single-tenant mode: separated runtime, data boundaries, and per-tenant configuration — built for enterprise expectations.

Sophona Architecture
Skill Map graph + runtime skills
Interaction LayerCore (Graph + Control)Execution & IntegrationsOne mind, many channelsPhone AgentVoice in/out · SIP/TwilioWeb WidgetEmbed · React/JSDesktop WidgetWindows/macOSAPI / iPaaSREST · Webhooks · QueuesAI Mind (Router)Interprets intent · selects path in the graphSkill Map Graph (Context Gate)Activates only relevant skills for the business pathMind InjectPolicies/procedures at runtime (global/path/skill)SchedulerCyclic · conditional · event triggersSkill Workflow RuntimeEach skill can plan · call AI · retry · verifyAutonomous WebClick · Navigate · FormsDesktop AutomationApps · OS UI · RPA-likeGoogle Workspace / APIsGmail · Calendar · CRM/ERP · iPaaSSkill Map is a graph used for context optimization: only relevant skills are exposed to the agent at a given time.
Sophona Runtime
Graph-gated skills + skill-level AI
User / TriggerVoice · Web · Desktop · APIAI Mind (Router)Intent · chooses graph pathSkill Map GraphContext gate: only relevant skills activeSelect Skill WorkflowPick skill node → run workflow runtimeSkill Runtime (AI inside)Plan · tool routing · retries · mind injectVerifierValidation · evidence · logsOutcomeAction done + responseMind Inject can apply globally, per path, and inside the skill runtimeGraph gating keeps context small. Skill workflows stay flexible: AI can plan, inject procedures, act, and verify — without hardcoded logic.

How

skills execute in runtime

The Skill Map selects the relevant branch (context gating). Then the chosen skill runs as an independent workflow that can plan, call tools, inject procedures, and verify outcomes. This makes behavior flexible, testable, and business-controlled.

01

Context gating via Skill Map graph

A trigger comes from phone / widget / desktop / API. The Skill Map graph activates only the subset of skills relevant to the current business path, optimizing context and reducing noise.

02

Skill workflow runtime (AI inside the skill)

The selected skill runs as its own workflow and can call AI live for planning, routing, retries, and tool usage — without bloating the global prompt.

03

Mind Inject at the right layer

Procedures can be injected globally, per path, or per skill. This provides predictable execution and full business control — without hardcoded decision logic.

Skill

Workflows

In Sophona, a skill can be tiny (like a single UI click) or a full, structured workflow with variables, conditions, retries and verification. This lets you choose between autonomy and determinism — without hardcoding agent logic. You control behavior by designing the skill graph.

Skills
Micro-skills ↔ Structured workflows
Spectrum: how “big” a skill can beMicro-skill (atomic)e.g. click / input / wait / read elementStructured workflow (stable)variables · conditions · deterministic stepsHybrid skilldeterministic flow+ AI only where neededMicro-skill building blocksUI.Click(selector)Low-level actionUI.Input(selector, value)Controlled inputUI.WaitFor(text/element)Sync / stabilityUI.Read(element)Extract evidenceStructured workflow skillsVariablesstate + parametersConditionsif/else branchesRetries & timeoutsno brittle flowsVerificationevidence + auditHybrid: stable + smartDeterministic steps control the process.AI is called only for:routing within the skillinterpreting unstructured inputchoosing next step when ambiguousgenerating a safe text responseKey idea: you can stay fully deterministic (no hallucinations) or allow autonomy — all by choosing which skills exist and how they connect.

Two ways to build skills

1) Micro-skills + autonomy

You can give the agent atomic skills like Click, Input, Wait, Read UI. Then the agent autonomously tries to complete the process by selecting and chaining these skills (guided by the Skill Map graph and Mind Inject).

Best for: changing UIs, exploratory tasks, “just get it done” operations.
2) Structured (stable) AI-powered workflows

Or you can build a classic workflow skill: variables, conditions, timeouts, retries, and verification. AI is called only where it adds value (interpretation, routing, drafting) — while the core execution stays deterministic.

Best for: compliance-sensitive flows, predictable outcomes, auditable execution.
No hallucinations by design

Reliability comes from structure: verification steps, evidence collection, and limiting AI to controlled surfaces. If you want 100% stable behavior, you can keep skills fully deterministic and use Mind Inject to enforce procedures.