Replay and audit AI agent runs.
The time machine for AI systems.
Replay and evidence infrastructure for AI agents. Capture executions as portable Run Capsules you own — then replay, compare, trace, and verify them. Self-hosted and local-first. Evidence you can replay.
$ pip install novafabricPython 3.12+ · Apache-2.0
the artifact
A trace is a view of a run.
A Run Capsule is an artifact you keep.
A NovaFabric Run Capsule is a portable execution-evidence artifact you own. Capture a run once; replay it, compare it with another, and — if you seal it with your own key — verify later that it has not changed.
- RUN
- 01HXAY7M5JZ8R7K4P9DPBYK2WX
- TIME
- 2026-10-08T09:14:02Z
- MODEL
- gpt-4o-mini · 6 calls captured
- TOOLS
- 3 tool calls captured
- HASH
- sha256:4f8a1c2e…9b07
- STATUS
- UNSEALED · SEALABLE
01 / THE SYSTEM
One fabric. Six domains of evidence.
NovaFabric is not a single tool — it's a coordinated system across six domains: capture, an evidence model, analysis, cryptographic trust, governance, and serving. Every piece composes.
Wrap any command. No code changes.
- Framework Adapters
- Multi-Target Runners
- Capture Orchestrator
- Wire-Level Hooks
- Secret Scanner
The portable, self-contained unit of record.
- Run Capsule
- Event Envelope v1
- Evidence Fabric
- Capsule Schemas
Replay, diff, trace lineage, evaluate.
- Replay Engine
- Diff Engine
- Lineage Graph
- Eval Suites
- Knowledge Graph
Seal with your key (opt-in), verify offline.
- NovaSeal Signing
- Evidence Bundle
- Merkle Log
Policy gates, approvals, compliance.
- Promotion (Maker-Checker)
- Policy Engine (OPA)
- Compliance & Export
- Governance & Judge
- Asset Registry
Multi-user API, dashboard, topology.
- Storage Layer
- Server (REST API)
- Dashboard
- Live Topology
Explore NovaFabric
Start with what you need.
Run capsules, replay, lineage, audit.
Open problems and the NovaFabric paper (arXiv:2609.12582).
Five composable building blocks.
Capture, seal, replay, diff, audit, across six domains.
Every version from v0.1 to current.
Community capsule gallery.
Getting started, concepts, CLI and API reference.
A walkthrough of capture, replay and diff.
The portable execution-evidence artifact you own: what is inside it and why it is not just a trace.
The five replay modes, and exactly what each one reuses and what it runs live.
Seal a run with your key and verify it offline months later.
How NovaFabric compares with Langfuse, LangSmith and OpenTelemetry.
paper/
NovaFabric: Tamper-Evident, Replayable Evidence for Autonomous AI Agent Runs
Mohsen Seyedkazemi Ardebili · arXiv:2609.12582, September 2026 · doi:10.48550/arXiv.2609.12582 · software archive: doi:10.5281/zenodo.22997388
Also: the technical overview, second edition, arXiv:2610.34594 (October 2026). See all publications.
What NovaFabric
is building.
Six active research and engineering areas, each labeled with its current maturity. Labels come from the NovaFabric source-of-truth docs and are updated only when evidence changes.
no capability uses stable until v1.0 schema freeze
Evidence + replay layer
Local-first run capsules for capture, replay, diff, validation, and audit workflows. CLI capture, SDK decorator, API/MCP proxy, and OTel GenAI semconv all captured.
NovaSeal trust layer
Tamper-evident DSSE/RFC 3161 sealing in local profile. Cryptographic SoD enforced. Enterprise HSM/X.509 PKI not claimed.
Lineage and provenance
SQLite lineage is experimental (10M-edge KuzuDB benchmark done). KuzuDB backend is prototype. Lineage federation is experimental with legal review pending.
Example runs.
Example capsules.
These are example capsules, not third-party deployments. Each snippet is read from the capsule files at build time; capture your own with nova capture.
submit your capsule →Open a PR on GitHub — add your capsule snippet to the showcase.
Standalone tools
in the ecosystem.
Behavioral test suite verifying object storage backends meet WORM requirements. 11 mandatory test cases covering immutability, retention locks, legal holds, lifecycle exemptions and checksum integrity. Tests object-store behavior against the requirements these rules describe; not a compliance certification.
Cluster-scale OpenTelemetry collector for capsule events, spans, and metrics. Deployable to Kubernetes via DaemonSet or SLURM via prolog/epilog. Requires Go 1.26.5+.
Web UI for capsule inspection, lineage visualization, and asset registry browsing. Ships embedded in nova serve. 30 tabs in 7 workflow groups (experimental).
Keep the run.
Keep the evidence.
AI agents call models, tools, files, and APIs across environments that keep changing. When an important run fails or is questioned later, the output alone is not enough to explain what happened.
NovaFabric captures those executions as portable Run Capsules so teams can replay, compare, trace, and verify past behavior without making a hosted dashboard the only record of what happened.
Open source. Apache-2.0. Self-hosted. No account required. No telemetry.
Portable by design
Run Capsules are portable directories designed to remain inspectable without a hosted service as the system of record.
Non-invasive capture
Supported SDK and transport hooks can capture agent executions without requiring application code changes.
Evidence-grade integrity
Sealing is opt-in: with your own key, a capsule gets a cryptographic signature (and, if you configure one, an RFC 3161 timestamp) so integrity can be checked later. NovaFabric provides evidence, not compliance certification.
Scale from laptop to cluster
SQLite for local development. Postgres + RBAC for multi-user server mode. Go collector for SLURM/Kubernetes cluster deployments.