DataShield Corpus v1.3.2
Corpus: the agentic skills MCP server
You ship agents to other people. Do you know which of them run a fork? DataShield Corpus, the MCP server for agent skills and docs, does. It keeps layered catalogs with drift detection for docs, agent definitions and skills. That means the vendor version, your changes, and their copies. All in one view.
Corpus is an agent-facing MCP server. It ships on the production stack. Auth meters and hash-chains every request.
An MCP server for agent knowledge with a layered catalog
Corpus serves the artifacts agents run on. That means documentation, agent definitions, skills, prompts and tool specs. All served over MCP. Auth meters and hash-chains every request. Corpus is not a dataset store. Governed data lives in Ontology.
On top sits the layered catalog:
- L0, the root catalog: the vendor's canonical documentation, agent and skill definitions.
- L1, the vendor layer: your changes and pinned versions. A patch layer over L0.
- L2, customer copies: each customer's deployed copy, diffed against L1 all the time.
Three layers, one diff. A customer forks when they edit a skill or drop a model pin. Corpus spots that drift. It names the drift too. You won't meet it for the first time in an incident review.
What you get from an MCP server for agent skills
Layered catalogs with drift detection
L0/L1/L2 catalogs hold vendor definitions, your changes, and customer copies of docs, agents and skills. You see them in one view, diffed all the time. The design is public as an RFC, and the code ships in Corpus.
Built on the production stack
Every Corpus request rides the shipped governance rails. That is Auth v2.4.1, with per-tool-call metering and a hash-chained audit. Governed data access stays in Ontology v2.2.0.
Early-adopter program
Early adopters get a direct say in the metering-schema roadmap. They also get a pricing lock at launch rates. Read the RFC first, then email support@myorg.ai.
Questions before you adopt Corpus
Can I use Corpus today?
Yes. Corpus v1.3.2 ships today. It is an MCP server for documentation, agent definitions and skills. It has the L0/L1/L2 layered catalog. Its usage metering federates into Auth.
How does Corpus relate to Ontology?
Ontology governs data. That is classification, tokenization, and analytics over tokenized datasets. Corpus governs agent knowledge. That is the documentation, agent definitions and skills your agents load. Corpus also tracks how the copies you shipped drift.
What does an early adopter actually get?
A say in the metering schema while the roadmap can still change, and a pricing lock at launch rates. In exchange, we get to hear what you need from drift detection early.
What is the difference between skills and MCP servers?
A skill packages expertise. It holds the instructions and files an agent loads to do a job your way. An MCP server is a live service. The agent calls it while it works, for tools and data. Corpus serves skills as versioned artifacts. Corpus is itself an MCP server too. Put simply: a skill tells the agent how. An MCP server gives it something to call.
Corpus v1.3.2 ships today. It federates with Auth and Ontology, both in production.
The shipped stack has pricing now, right here: Auth, Ontology, Commander and Corpus.
See pricingYou've seen the proof
Ready for a number? Scope your deployment and we'll price it against your own economics.
Get your quote →