14 , advanced
Advanced
Advanced capabilities for production-scale governance including chaos testing, fleet management, cross-org federation, operator governance, provenance tracking, scheduled executions, and compensation workflows.
Chaos testing
Faramesh ships with 22 built-in chaos scenarios that simulate backend failures, latency spikes, and policy edge cases. Use them to verify that your agent degrades gracefully under adverse conditions.
# List all available scenarios
faramesh chaos-test list-scenarios
# Run a specific scenario
faramesh chaos-test run --agent support-bot --scenario prompt-injection
# Toggle degraded mode on the daemon
faramesh chaos-test degraded toggle
Fleet management
Manage policies and kill switches across all agents in a fleet. Push policy updates atomically and monitor compliance across deployments.
# List all agents in the fleet
faramesh fleet list
# Push a control event to one instance
faramesh fleet push instance-01 --message "policy rollout"
# Kill a specific instance
faramesh fleet kill instance-01 --reason "security incident"
Cross-org federation
When agents from different organisations interact, federation provides mutual governance verification. Each organisation trusts the other's Faramesh instance and exchanges signed governance receipts.
# Add a trusted external organisation
faramesh federation trust add \
--org partner-corp \
--bundle partner-trust-bundle.pem \
--scope "*"
# Verify a governance receipt from a partner
faramesh federation receipt verify <receipt-token>
Operator governance (four-eyes)
Policy changes require two-person approval. One operator proposes the change; a second operator approves or rejects it. All proposals are recorded in the audit trail.
# Propose a policy change
faramesh ops policy-change propose \
--policy policy-v2.yaml \
--reason "Add refund ceiling rule"
# Second operator approves
faramesh ops policy-change approve <proposal-id>
Provenance and model integrity
Sign and verify policies and model artifacts. Provenance records form an independent integrity chain that can be audited alongside the DPR chain.
# Sign agent provenance
faramesh provenance sign --agent support-bot --framework langgraph --model gpt-4o --tools "search,db" --key ./signing-key.pem
# Register and verify a model
faramesh model register gpt-4o --provider openai --version 2026-03 --fingerprint sha256:abc123
faramesh model verify --agent support-bot
Scheduled executions
Schedule agent executions with policy-governed approval. Schedules can require manual approval before firing.
# Create a scheduled execution
faramesh schedule create \
--tool report/generate \
--args '{"window":"24h"}' \
--agent report-bot \
--at "2026-01-15T09:00:00Z" \
--reeval
# Approve a pending schedule
faramesh schedule approve <schedule-id>Compensation workflow
When a multi-step agent action partially succeeds, the compensation workflow provides structured rollback. Build a compensation plan that pairs each forward action with its undo operation.
# Build a compensation workflow
faramesh compensate build <record-id> \
--policy policy.yaml
# Apply compensation on failure
faramesh compensate apply <compensation-id>
# Retry a failed compensation step
faramesh compensate retry <compensation-id> --from-step rollback-payment