Create a mock LangChain-like chain/runnable.

Add an agent with its dependencies

Calculate trust score from 0-1000. Base score from verification level + modifiers from behavior and capabilities.

Initialize .agents/ directory with Agent OS support.

Create a new file in the file system. Args: path: Path where to create the file content: Initial con

Execute a runbook's steps sequentially against an incident. Args: runbook: The runbook to execute. incident: The inci

Create a failed result.

Create episode from dictionary.

Open the pipe for communication.

Snapshot for the ``pre_tool_call`` intervention point (§2.5).

Reconstruct a Decision BOM from observability signals. Provide either a trace_id (preferred, most precise) or an agent_id + timestamp

Get file metadata.

Add a step to the pipeline.

Calculate drift/hallucination score between two outputs. This is the primary verification function - a pure function with no side effects.

Verify hash-chain integrity of the entire log.

Calculate drift score between two embedding vectors. Enhanced verification function with configurable metrics, weighting, threshold prof

Verify hash integrity of a memory entry. Returns ``True`` if the stored hash matches a fresh computation.

Verify a peer before communication.

Write a single entry, rotating the file if necessary.

GenerateIdentity creates a new Ed25519-based identity for the given agent.

Log a message with agent context.

Create a TrustAwareAgent with a pre-configured trust store.

Create a mock RunHooks context.

Activate the agent (from provisioning, suspended, rotating, degraded, or quarantined).

Add a capability grant to this scope. Args: grant: The ``CapabilityGrant`` to add. Raises: ValueError: If th

Calculate current burn rate within a time window. Burn rate = (actual errors in window / window size) / (budget / total window) A bur

Check if an agent is trusted enough to call a function. Order of checks: 1. Function not blocked 2. Trust score meet

Check if an action violates the constraint graph. Uses a three-level check: 1. Role-Based Check: Is this tool allowed for th

Connect to the broker.

Return ASGI middleware for LiteLLM proxy JSON calls. This adapter targets the LiteLLM proxy server app. Install proxy support with ``pip inst

Check whether an agent is permitted to perform an action.

Guard tool invocations against capability policy.

Process an Agent OS audit log entry and return a Signal if relevant.

Return True if *alert* is novel and should be delivered. .. note:: For multi-threaded delivery, prefer :meth:`consume`, which

Spawn an agent function in an isolated process.

Create a hardened Docker container for the session. Args: agent_id: Validated agent identifier. session_id: Internall

Shortcut: create a detector and scan a single input.

Create a KernelMCPServer, skip if not installed.

Create a mock PydanticAI tool entry.

Set up a FederationEngine with pre-loaded data.

Assign a role to an agent.

Get a scanner instance by name.

Return a fresh :class:`AdapterRuntimeBridge` for ``policy``. Adapters call this once per kernel instance. Pass ``runtime`` to inject a pre-bu

Handle MCP tools/call request.

── K8s ──────────────────────────────────────────────────────────────────────

A ready-to-use provider with the default sandbox queued.

Receive the next message. Args: timeout: Maximum seconds to wait. None means wait forever. Returns: Received

Record a reward signal for an agent. Args: agent_did: The agent's DID dimension: Which dimension this affect

Record a tool invocation and check for anomalies.

Validate a value against all validators (and the AGT bridge if wired). Returns a ValidationResult with aggregated outcomes and the a

RecordSuccess increases an agent's trust score with decay.

Helper to create a plugin with the given code.

Consult the PolicyEvaluator if one is configured. Returns ``(allowed, reason)``. When no evaluator is set, returns ``(True, "")`` so

Get current UTC timestamp in ISO 8601 format.

Get the global Hypervisor instance.

Generate a fresh Ed25519 agent public key.

Run a demo script and return its combined stdout+stderr.

Create a plugin subdirectory with an agent-plugin.yaml manifest.

Add a step to a saga.

Cancel the task.

Classify the risk level for an AI system per EU AI Act Article 6. Evaluates the profile through the following cascade: 1. **Article

Check whether a plugin manifest complies with a marketplace policy. Args: manifest: The plugin manifest to evaluate. policy: The

Build a public pre-key bundle for distribution. Args: otk_id: Optional one-time pre-key ID to include.

Get the current Merkle root hash.

Check whether this endorsement has passed its expiry time.

Load policy from a YAML file.

Log action to audit trail.

Record an action for regime detection.