Sitemap

Synedrix OS - Site Map

Every page of the site, grouped by topic. The same list is available as JSON at /sitemap.json and as XML at /sitemap.xml.

Product

Synedrix OS - The governed operating layer for agentic business systems.
Platform - A reusable, self-hostable platform for governing agentic business execution: cloud-connected, self-hosted hybrid, or air-gapped.
Architecture - The twelve canonical layers of Synedrix OS, in depth: what each one owns, what it must not own, and how it's implemented.
Operating model - How Synedrix OS separates human accountability from machine-coordinated execution: four organisational planes, the human control roles, and how work flows.
Principles - The 12 principles that constrain how Synedrix OS behaves, and the enforcement mechanism behind each one.
Domain packs - How Synedrix OS stays reusable across industries while each domain contributes what is genuinely domain-specific.

Architecture layers

Layer 1: Human Interface - Where a person meets the system. It renders and routes; it does not decide.
Layer 2: Goal & Intent - Turns what a business wants into something the system can plan against.
Layer 3: Orchestration - The center of gravity. It sequences and retries; it does not reason.
Layer 4: Agent Runtime - Where reasoning actually happens. It loops, accumulates state, and resumes.
Layer 5: Agent Registry - The gate every invocation passes through. It says who may run, not what to do.
Layer 6: Tool Abstraction - One tool contract, one authorization model, regardless of what's behind the adapter.
Layer 7: Memory & Data - Five Layer 7 stores and one governed lifecycle. Nothing moves without leaving lineage.
Layer 8: Event - One envelope for every fact in the system, from ingress to egress.
Layer 9: Governance & Controls - Decides whether, never performs. Every execution path passes through it first.
Layer 10: Observability - What happened, how well, and can it be reconstructed, for every layer, not a sample.
Layer 11: Domain Pack - Where the OS stops being domain-agnostic, and one business's logic begins.
Layer 12: Business Instance - Where a reusable definition becomes one customer's actual deployment.

Platform components

n8n (Workflow kernel) - The orchestration kernel. It sequences and retries; it does not reason.
LangGraph (Reasoning engine) - Where reasoning actually happens. It loops, accumulates state, and resumes.
Hookdeck (Ingress gateway) - The queue at the edge. Events arrive whether or not the orchestration layer is up.
Hookdeck Outpost (Egress gateway) - The last observable hop. Delivery guarantees as infrastructure, not workflow logic.
Redis (Coordination bus) - Safe retries and rate limits come from Redis's own primitives, not from features built on top.
Supabase PG17 (Relational system of record) - Canonical relational truth. It is Postgres, undisguised.
QuestDB (Time-series engine) - The time axis as a first-class citizen, queried in ordinary SQL.
Neo4j (Lineage graph) - Lineage is a traversal problem. Nothing else in the stack answers reachability natively.
Qdrant (Vector memory) - Answers what is similar. Filtering happens inside the search, not around it.
Langfuse (LLM engineering) - One observability plane across orchestration and runtime. Traces are not an audit trail.
PostHog (Product analytics) - Is the product working for the people using it? A different question from system health.
Grafana LGTM Stack (Operational monitoring) - Is the system healthy? Five projects composing into one observability backend.
LanceDB (Embedded vector index (EverOS)) - The embedded vector store EverOS ships with: evaluated in isolation, never in the stack.
EverOS (Local-first agent-memory runtime) - Its memory lifecycle was already adopted as canon. Its runtime (Markdown, SQLite, LanceDB) was not.
Semantica (GraphRAG / knowledge-graph toolkit) - Structural parsing and decision-graph tooling, evaluated one capability at a time, not one adoption question.
Docling (Structural document parsing) - The one concrete, testable claim in Semantica's whole capability bundle: table-aware parsing, measured against the existing chunker.
Laya (Decision model, self-hosted) - A model that makes bounded decisions without writing a word, and runs wherever Synedrix OS runs, air-gapped included.
Jev (Decision model, managed API) - The managed reference point: the quickest way to measure what a decision model can do, and never the default.

Learn and explore

Academy - What the Synedrix OS Academy is, why it exists, and who it teaches.
Labs - A sandbox for new technology: real tools, run for real, in a sandbox that can't touch production.
Engineering Room - A weekly note on what the team worked on while building Synedrix OS: what moved, what is still in progress, and what comes next.
Governed Agentic Development Lifecycle (GADLC) - How Synedrix OS is built: AI agents working under human direction, held to the same five obligations the platform applies to any agent.

Blog

Synedrix OS blog - Articles on the Synedrix OS architecture, governed agentic systems, and the work behind building them.
The authority boundary: what agents can and cannot do - An agent's authority belongs in a versioned record, not its prompt. See how Synedrix OS uses recommend, approve and execute scopes to bound what agents do.
Governance before execution: why policy is an OS primitive - Synedrix OS checks proposed actions before they run. Learn how policy decisions, human approvals and business thresholds govern a refund through execution.
Memory isn't a bucket: how Synedrix OS separates kinds of truth - Synedrix OS separates kinds of memory, the stores that hold them and who may read or change them, so agents use information with clear permissions and evidence.
The 12 layers of Synedrix OS: a complete mental model - Each of the 12 Synedrix OS layers states what it owns and what it must not own. See all twelve in five slices and why the exclusions make drift detectable.
Why specialized agents beat mega-agents every time - Specialised agents give tasks, permissions and results clear owners. See how Synedrix OS coordinates their work and makes errors easier to test and investigate.
Orchestration as the center of gravity - Synedrix OS puts coordination and delegation at the centre so agents have clear responsibilities and their work can be understood when something goes wrong.
The 12 principles that make agentic systems trustworthy - Twelve design principles in five clusters set the limits that make an agentic system trustworthy. See what each requires and why trust is the aim they serve.
One substrate, many domains: the domain pack model - Domain packs let one governed platform serve trading, commerce and operations. See how the rule of inclusion decides what the OS, a pack and a company own.
Introducing Synedrix OS: the governed operating layer for agentic business - Synedrix OS is the governed operating layer for agentic business systems: routing, agent lifecycle, shared context, governance and observability built once.
From static workflows to agentic operating systems - Static workflows encode a fixed process and push every exception to a person. See how five capabilities coordinate specialised agents toward a goal instead.
Why AI assistants don't solve the coordination problem - AI assistants can speed up individual tasks while approvals and handoffs stall. Learn what coordination across people and systems requires beyond faster work.
The coordination debt that's quietly costing enterprises their edge - Coordination debt is the cost of people translating context across enterprise systems. Learn why new tools speed tasks but leave organizational handoffs intact.

Blog topics

Governance articles - How Synedrix OS checks proposed actions before they run, with policy decisions, human approvals and business thresholds that keep execution accountable.
Agents articles - How specialised agents in Synedrix OS take on defined roles, hand work to each other and stay testable, keeping recommendations, approvals and actions apart.
Architecture articles - How Synedrix OS divides a governed agentic system into twelve layers, what each layer owns and must not own, and how the boundaries keep changes reviewable.
Orchestration articles - How Synedrix OS turns a goal into delegated work: dividing tasks, assigning agents, handling failure and keeping the evidence that explains each result.
Coordination articles - Why coordination between people, agents and systems is the real constraint on agentic work, and what shared context, clear ownership and handoffs require.

Engineering Room

Week 45 · 2–8 Nov 2026 - A week spent on review and publishing automation, a move to a new organisation account, and topics and shared code panels for the blog.
Week 44 · 26 Oct–1 Nov 2026 - A week of publishing and review automation, with the first blog series going out and the specifications turning to event handling.
Week 43 · 19–25 Oct 2026 - An architecture week on agent accountability and executable manifests, with the website's forms moved onto a governed workflow.
Week 42 · 12–18 Oct 2026 - A week on how documentation is approved for release, with new design proposals for the client application, process governance and manifests.
Week 41 · 5–11 Oct 2026 - The website was prepared for launch, the specification gained a twelfth principle, and content gained a lifecycle and shared phrasing.
Week 40 · 28 Sep–4 Oct 2026 - The specification took on tenancy, identity and metering, plan governance gained its specification stage, and the website was checked for launch.
Week 39 · 21–27 Sep 2026 - The website explained the operating model in plain language, and a new repository began governing how plans are proposed and checked.
Week 38 · 14–20 Sep 2026 - A quiet week until the weekend, when the first blog drafts were grounded in the specification and the content repository took editorial charge.
Week 37 · 7–13 Sep 2026 - The teaching material grew its first three chapters, and the specification defined how goals, task graphs and planning fit together.
Week 35 · 24–30 Aug 2026 - A short week. On Monday we explored a domain pack for a developer assistant that summarises work in progress.
Week 34 · 17–23 Aug 2026 - The chat agent gained an evaluation harness, a governed lane for lab evaluations opened, and every repository took the same branch and PR checks.
Week 33 · 10–16 Aug 2026 - The documentation site was rebuilt over the source repositories, the chat agent gained tracing, and the specification described each platform.
Week 32 · 3–9 Aug 2026 - A short week. The agent registry got its first document, and the reference stack's model settings were tuned to avoid repetitive output.
Week 31 · 27 Jul–2 Aug 2026 - The website and the shared design system began, and the specification gained a machine-readable facts layer checked on every change.
Week 29 · 13–19 Jul 2026 - A quiet week. The only change kept the reference stack's weekly overview refresh up to date.
Week 28 · 6–12 Jul 2026 - Each repository gained a generated overview refreshed every week, and the reference stack took the proof of concept's database schema.
Week 27 · 29 Jun–5 Jul 2026 - A first proof of concept connected an agent to the workflow engine, with its contract and a specification for tool authorization.
Week 26 · 22–28 Jun 2026 - A quiet week until Sunday, when the specification gave the agent runtime its own layer, separate from orchestration.
Week 25 · 15–21 Jun 2026 - A quieter week with operating guidance for the vector database and the eighth post in the blog series.
Week 24 · 8–14 Jun 2026 - The model gateway gained tracing and several open-weight models, and laptop models joined it behind one entry point.
Week 23 · 1–7 Jun 2026 - The self-hosted stack moved onto an NVIDIA DGX Spark, serving language models locally behind a model gateway.
Week 22 · 25–31 May 2026 - A quieter week on the self-hosted stack, prepared for an NVIDIA DGX Spark with adjusted databases and dashboards.
Week 21 · 18–24 May 2026 - A broad design week covering new domain pack designs, how agentic workflows run and report, and a home for content production.
Week 20 · 11–17 May 2026 - The specification took shape as linked documents covering observability, signals, integration and the stack, plus a domain pack template.
Week 19 · 4–10 May 2026 - Work resumed at the end of the week with the groundwork for the documentation, the architecture, a naming convention and a consolidated set of documents.
Week 17 · 20–26 Apr 2026 - Synedrix OS got its own specification, and the self-hosted stack gained backups, health checks, log collection and its first dashboards.
Week 15 · 6–12 Apr 2026 - A quieter week on foundations, in which a documentation repository was set up and the self-hosted workflow engine was updated.
Week 14 · 30 Mar–5 Apr 2026 - Agent registry, design principles and database schema written down, with human approval built in, and a repository for the self-hosted stack.
Week 13 · 23–29 Mar 2026 - The first week of the project, in which a repository for agentic workflows was created and given its first description.

Company

About - Why Synedrix OS exists, who is building it, and where it stands.
Why Synedrix - The name comes from the Greek synedrion, a sitting together: what the word means and why it describes a governed team of AI agents.
Invest - An early-stage, self-hostable AI infrastructure platform with a complete architecture. Early investment conversations are open.
Build with us - A complete architecture and an active build underway. The founding team is forming, and co-founders are welcome.
Early access - Follow the build: a weekly update on the development in progress, with a sneak preview, and a notice at launch.
Contact - Three ways to engage with Synedrix OS: follow the build, build with us, or invest.
Legal - Privacy policy and terms of service.

Latest post

The authority boundary: what agents can and cannot do

An agent's authority belongs in a versioned record, not its prompt. See how Synedrix OS uses recommend, approve and execute scopes to bound what agents do.

Read