System Architecture
The Parcle Company Brain connects company data, employee devices, and agents into one governed platform. Agents can reach the sources they are permitted to use.
The Company Brain is a single platform with two kinds of inputs and one governed core. On one side, your company's data sources flow in through connector-linked tools and structured data platforms. On the other side, your employees' devices and the agents running on them connect in: they both contribute memory and consume knowledge. Everything meets in one brain that every agent queries through a single, access-controlled layer.
The platform at a glance#
Two kinds of inputs#
Connectors bring in SaaS tools and documents (Google Drive, Notion, GitHub, HubSpot, Linear, and 30+ more). Structured data platforms (Snowflake, Databricks, BigQuery, Postgres) are connected as live data sources. Parcle reads them in place rather than copying the data. See Connectors.
Employee and developer machines are enrolled as managed devices via Team Sync. The agents on them, including Codex, Claude Code, Openclaw, Cursor, or Hermes, connect to the brain to write memory (what they learn and are told) and query it for grounded knowledge.
The two sides meet in one store. Data sources build the connected-knowledge half of the brain; devices and agents contribute the shared-memory half. A single query and governance layer serves both. That is what lets any agent reach any source it is permitted to.
Inside the brain#
The platform core is structured as four layers, wrapped by a governance plane:
| Layer | What it does |
|---|---|
| Ingestion & Sync | Accepts connectors, documents, and agent memory; reads your sources and infers concepts. Ingestion runs in the background, so nothing blocks while content is processed. |
| Memory & Storage | Holds the company's shared memory and connected knowledge in one place. See Memory & Knowledge. |
| Ontology Graph | Links factual nodes (your structured data and documents) to concept nodes (business terms), forming the semantic map that turns plain-English questions into precise answers. See Ontology Graph. |
| Query & Reasoning | Traverses the graph, executes the needed steps (queries, lookups), and streams back grounded answers with citations and a confidence signal. See Data API. |
| Governance & Access Control | Authenticates every request and applies company permissions, roles, SSO, and audit across everything above. See Security. |
Unified, but governed#
The platform's value is that it collapses N agents × M sources of bespoke integration into one connection plane and one memory plane. But "company-wide" never means "everyone sees everything":
Every agent sees only the knowledge and memory it has been granted access to, defined by company permissions and roles.
Access rules are defined once, centrally, and enforced for every agent and every source, not re-implemented per agent.
Every query and answer is recorded; SSO ties access to your company's identity provider. See SSO & Audit.
So an agent reaches "any source" through the brain only within the permissions it has. The platform is the single place those permissions are defined and enforced.
For where each part of the platform can run, including managed cloud, isolated single-tenant, or sandboxed / self-hosted within your perimeter, see Deployment Models.
Where the platform can run — managed cloud, isolated single-tenant, or self-hosted within your perimeter.