Praesidia Documentation
Connect your first workflow, configure its controls, and check the result. Follow one agent action through its evidence and verification, or choose a starting path for agent access and model spend.
Quickstart
Getting Started
Get up and running with Praesidia in minutes. Register your first entity, create a connection, and secure your AI infrastructure.
Your First Model Request
Configure a supported model provider, send one OpenAI or Anthropic request through Praesidia, and inspect the recorded usage.
From Agent Action to Audit Evidence
Connect a managed MCP action to its authorization record, event timeline, signed audit export, and independent verification.
Install a matching runtime SDK
Install the Praesidia SDK release your Praesidia contact provides, and separate runtime credentials from personal evidence review.
Concepts
Entity Types
Learn about the three entity types in Praesidia: Applications, MCP Servers, and Agents.
Connections
Learn how connections work in Praesidia. Define trusted relationships between entities with security controls.
Guides
Guardrails
Configure content-level controls on connections between AI entities. Guardrails define what can be communicated.
Policies
Configure operational controls on connections. Set rate limits, geographic restrictions, time-based access, and more.
Evaluation Datasets
Build named, versioned golden datasets for agent evaluations: add cases by hand, sample them from production traffic, and compare regression runs.
AI Agent Risk Register
Record EU AI Act Article 9 risks per AI system, asset or agent: category, mitigations, residual risk, eight risk dimensions, incident link and review date.
EU AI Act Risk Classification
Classify an agent, MCP server, or application under the EU AI Act's risk tiers, track compliance status and remediation, and get re-assessment reminders.
Governance Packs
Install a pre-built bundle of guardrails, intent rules, and compliance-risk mappings for a regulated industry, then adjust it for your organization.
Trust Scores and Agent Drift Detection
How Praesidia scores agent trust from multiple signals, accepts signed third-party attestations, and detects an agent drifting from its own baseline.
OpenClaw: setup and scope
OpenClaw on Praesidia: limited coverage. Native managed tool, owner-context checks and strict before-tool blocking pass pinned plugin acceptance.
Hermes Agent: setup and scope
Hermes Agent on Praesidia: limited coverage. Python managed HTTP tool is implemented with durable approval and exact-request recovery.
ZeroClaw: setup and scope
ZeroClaw on Praesidia: limited coverage. Authenticated instance management, durable delivery and a native restricted MCP tool profile are implemented.
NemoClaw / OpenShell: setup and scope
NemoClaw / OpenShell on Praesidia: limited coverage. A version-matched native tool gate and managed MCP recipe are implemented; full NVIDIA sandbox.
LangGraph / LangChain / Deep Agents: setup and scope
LangGraph / LangChain / Deep Agents on Praesidia: limited coverage. Python managed HTTP tool is implemented with durable approval and exact-request recovery.
CrewAI: setup and scope
CrewAI on Praesidia: limited coverage. Managed tool passes functional fixtures; release is blocked by unresolved upstream Chroma dependency advisories.
OpenAI Agents SDK: setup and scope
OpenAI Agents SDK on Praesidia: limited coverage. Separately versioned Python and TypeScript managed tools with native approval callbacks and durable restart.
Google ADK: setup and scope
Google ADK on Praesidia: limited coverage. Python managed HTTP tool is implemented with durable approval and exact-request recovery.
Microsoft Agent Framework: setup and scope
Microsoft Agent Framework on Praesidia: limited coverage. Python managed HTTP tool is implemented with durable approval and exact-request recovery.
Agno AgentOS: setup and scope
Agno AgentOS on Praesidia: limited coverage. Python managed HTTP tool is implemented with durable approval and exact-request recovery.
OpenCode: connection and validation
OpenCode on Praesidia: limited coverage. OpenCode 1.18.29 authenticates and discovers the runnable managed MCP companion. Explicit prepare, owned checkpoint.
Claude Code: connection and validation
Claude Code on Praesidia: limited coverage. Claude Code 2.1.202 authenticates and discovers the runnable managed MCP companion. Explicit prepare, owned.
n8n: connection and validation
n8n on Praesidia: limited coverage. Installable n8n node 0.1.0 executes authenticated managed MCP prepare, checkpoint and explicit approved resume against an.
Dify: connection and validation
Dify on Praesidia: limited coverage. The installable tool plugin passes actual Dify Plugin SDK 0.10.2 loading and authenticated managed MCP prepare.
Langflow: connection and validation
Langflow on Praesidia: limited coverage. The runnable component passes actual lfx 1.12.0 invocation over authenticated managed MCP, including prepare, owned.