Documentation

Agent Context

Raw file sync to your agent's filesystem. Give agents direct access to XML schemas, CSV datasets, PDF manuals, and any file they need to do their job.

“Context gives your agent a filesystem. KB gives your agent a brain. Combined, your agent knows HOW to work AND has the raw data to work with.”

What is Context?

Context is the second layer of domain knowledge an agent receives — alongside the Knowledge Base (Neo4j graph queries via MCP). While KB contains queryable “know-how” documentation, Context gives agents raw files directly on disk.

Filesystem Access

Files are synced directly to your agent's local filesystem at pod startup. Agents read, parse, and reference them like any local file.

Any File Format

XML schemas, CSV datasets, PDF manuals, JSON configs — any file your agent needs. No conversion or re-formatting required.

Multi-Source Sync

Pull files from S3 buckets, GCS buckets, or Git repositories. Sources are synced automatically when the agent pod starts.

Zero-Config Access

No MCP tools or API calls needed. Files appear at a predictable path — agents just open and read them directly.

Combine with KB

Use Context for raw files and KB for queryable knowledge. Together, your agent has both structured know-how and raw reference data.

Dashboard Config

Add, remove, and manage sources from the dashboard. Each source maps to a folder on the agent's filesystem.

Agent Filesystem

/home/appuser/
├── workspace/              ← agent working directory
└── context/
    └── <source-name>/      ← synced files organized by source
        ├── schema.xml
        ├── reference.csv
        └── manual.pdf

Source Types

When you add a source, you choose how the data is delivered to your agent. Each type serves a different purpose.

type: kb

Knowledge Base

Documents are ingested into the Neo4j knowledge graph. Agents query them via MCP tools (semantic search, graph traversal).

Accessed via MCP tools

type: context

Context (Files)

Raw files are synced to the agent's local filesystem. Agents read them directly — no API needed.

/home/appuser/context/<source-name>/

type: both

Both

Documents go into the Neo4j graph AND raw files land on disk. The agent gets queryable knowledge plus the original files.

MCP tools + /home/appuser/context/<source-name>/

Getting Started

Set up file sync for your agent in three steps.

Add a source

Go to the Context page on your dashboard. Click "Add Source" and provide credentials for your S3 bucket, GCS bucket, or Git repository.

Get started free

Choose source type

Select "context" for raw file sync, "kb" for knowledge graph ingestion, or "both" for the full experience. Each source gets its own folder.

Agent reads files

When your agent starts, files are synced to its filesystem automatically. The agent can open, parse, and reference them during any task.

Real-World Example: ERP Coding Agent

An ERP software company uses both KB and Context to build a domain-expert coding agent:

KB (Know-How)

OFBiz process documentation, coding patterns, and best practices. The agent queries these via MCP to understand how to build ERP features.

Context (Raw Files)

365 OFBiz XML entity definition files on disk. The agent reads and parses actual schema structures to generate correct code.

Result: An agent that understands ERP concepts AND can read/parse actual XML schemas — producing code that matches the real data model.

Ready to give your agent a filesystem?

Add your first source and let your agent access the raw files it needs to excel at domain-specific work.