§ 01 — Modern Relay

Context &
Coordination
Infra

The governed foundation for teams and agents to share context, coordinate work, and compound knowledge.

Our products GitHub

[Context & Coordination]

Solutions for agent-native teams

Company Brain

Problem

Company knowledge is scattered across wikis, tickets, Slack and a dozen SaaS tools. No team or agent sees the whole picture, and institutional memory walks out the door with the people who hold it.

Solution

One governed graph consolidates docs, decisions, tickets and systems of record, so every team and every agent reads from and writes to the same typed source.

Key pillars

Single source of truth

one graph, not a dozen silos

Ontology

a typed semantic layer, not opaque blobs

Governed writes

every change reviewed and attributable

Context Graph

Problem

Agents lose the thread on long-horizon work. They forget prior decisions, repeat each other's mistakes, and overwrite each other's state with no record of who changed what or why.

Solution

A shared, versioned graph holds decisions, tasks and their dependencies. Each agent works on its own branch, and changes merge only after review.

Key pillars

Versioned memory

a branch per agent or task

Connected work

tasks, code and decisions linked, not siloed

Durable context

state survives across runs and handoffs

GraphRAG

Problem

Pure vector search retrieves text that looks similar but misses how facts connect. Answers drift, hallucinate, and can't be traced back to a source.

Solution

A single query traverses typed relationships and ranks results with vector, full-text and BM25, returning answers that cite the exact nodes behind them.

Key pillars

Hybrid retrieval

vector, full-text and BM25 in one runtime

Structure-aware

traversal over typed edges, not embeddings alone

Cited answers

every result traces back to its nodes

On-Prem AI Infra

Problem

Regulated and security-conscious teams can't send proprietary data to a SaaS vendor, and most AI platforms are thin cloud wrappers that assume your data leaves the building. Residency, compliance and lock-in concerns stall projects before they start.

Solution

The full Modern Relay stack runs inside your own environment: the context graph, the workspace and the agent surfaces, deployed on-prem, in your VPC, air-gapped or hybrid. It connects to the systems and models you already run, on any S3-compatible store, and your data never leaves your network.

Key pillars

Full stack, not a wrapper

graph, workspace and agents in one deployment

100% in your environment

on-prem, VPC, air-gapped or hybrid

Your data, your models

connects to the systems and models you already run

Browse the cookbook

[State of Affairs]

Intelligence is here. Systems are not ready.

Every system still assumes a human.

CRMERPTicketsWorkflows

1 Operatorread · judge · approve

One operator behind every screen.

The operator changed.

Agents act as operators.

They don't just ask questions. They deliver real work:

Reach across systems
Update records
Trigger workflows
Hand off tasks
Write and execute code

One agent is easy.

How about dozens?

parallel · same systems · partial context · overlapping decisions

Without coordination, they collide.

✕✕✕

overwrite · stale reads · no audit trail

Who changed what? No one knows. No one knows why.

Today, teams reach for tape.

connectorspluginsscriptsglueworkflows

Every new agent adds another wire. The system gets more fragile.

Agents need a substrate.

read
branch

write
merge

Treat knowledge like code. It doesn't exist yet — so we're building it.

Read the manifesto

[Primitives]

Building blocks

Not a monolithic SaaS that locks you in — composable building blocks that seamlessly assemble into your existing infrastructure.

Data Branches

Git-style data branches — agents propose changes in isolation, reviewed and merged like code.

Graph

Typed nodes and edges — semantics and ontology.

Surfaces

Slack, VS Code, CLI, web, Claude — meet the graph where you already work.

Policy

Access and governance, defined as code.

Ontology

A model of your domain. A shared contract for agents and humans.

MD File

The native format. Human-readable, agent-editable.

ETL Pipeline

Deterministic and reproducible ingestion.

Object Storage

S3 — infinitely scalable storage at the lowest cost per GB.

Declarative Config

Code defines everything about your system. Consistency, auditability, reproducibility.

Events

Event streaming across the graph.

Notebook

Declarative UIs for teams.

Open Table Format

Lance as open lakehouse format for multimodal data.

Omnigraph orchestrates these primitives

[Product]

Omnigraph

The lakehouse-native graph engine for context assembly. The operational state and coordination substrate for agents.

A typed knowledge graph that lives directly on object storage: versioned, branchable and queryable. Agents read from it, write to it, and coordinate through it.

RustLanceS3Cedar

Git-style

Versioning, branches, commits and merges over a typed graph.

Multimodal retrieval

Graph, vector, full-text and filters across text and blobs.

Object-storage native

Runs on S3 or RustFS. Elastic storage.

Open format

Lance as the open storage layer — read it with anything.

as code

Schema

Typed .pg schemas — planned, applied, enforced.

as code

Context

Linted queries and agentic nudges, versioned and reusable.

as code

Security

Cedar policies enforced server-side on every mutation.

as code · soon

Dashboards

Declarative views and controls over the graph.

Try Omnigraph

[Deployment]

Deployed on infrastructure you own

Modern Relay's stack is optimised for data sovereignty and maximum infrastructure control.

Option 1

On-prem

Fully inside your own data center.

Runs on your hardware, your network
Air-gap friendly — no outbound calls
Object storage on local disk or on-prem S3
You hold every key

Option 2

VPC

In your own cloud account.

Deploys into your AWS, GCP or Azure VPC
Your bucket, your IAM, your peering
Private networking — nothing leaves your tenancy
Scales with the cloud you already run

Option 3

Hybrid

On-prem and cloud, one graph.

Keep sensitive data on-prem, burst to cloud
Branch and sync across environments
One control plane spanning both
Move workloads without re-platforming

Read the deployment guide

[Writing]

Field notes

Read all memos

[Careers]

Build with us

We hire high-agency people, genuinely passionate about data engineering, knowledge graphs, ontologies and infra for agents. Small team, high bar, real ownership.

View all roles & details

[Investors]

Context &CoordinationInfra

Solutions for agent-native teams

Company Brain

Intelligence is here. Systems are not ready.

Building blocks

Data Branches

Graph

Surfaces

Policy

Ontology

MD File

ETL Pipeline

Object Storage

Declarative Config

Events

Notebook

Open Table Format

Omnigraph

Git-style

Multimodal retrieval

Object-storage native

Open format

Schema

Context

Security

Dashboards

Deployed on infrastructure you own

On-prem

VPC

Hybrid

Field notes

The Knowledge-Coordination Manifesto

Agents Are the New Database Operator

The Branchable Context Graph: A New Operational State Layer for Agentic Systems

Build with us

Founding Engineer, Data & Storage

Founding Developer Advocate

Backed by

Context &
Coordination
Infra