| What it does | The model training platform for AI teams
| See, trace, stop, and prove what your AI agents do
|
|---|
| Stage | Not given
| Startup
|
|---|
| Founded | Not given
| 2026
|
|---|
| Based in | Not given
| AU
|
|---|
| Pricing model | Freemium
| Freemium
|
|---|
| Pricing | Free: $0/month with $50 in credits once, 1 seat per project, every feature, console, CLI and MCP server, community Discord support. Pro: $120/month with $120 in credits every month, unlimited seats, priority support. Enterprise: Custom pricing, contact for details.
| Free: Community edition (1,000 runs/month, 1 workspace, 7-day retention, self-hosted). Starter: A$49/month (20,000 runs/month, 5 seats, 30-day retention). Growth: A$600/month (50,000 runs/month, 10 seats, 60-day retention). Scale: A$1,800/month (75,000 runs/month, 20 seats, 60-day retention). Pro: A
|
|---|
| Key features | - OpenTelemetry agent tracing and trace import from Langfuse, LangSmith, Braintrust and Galileo
- Generated evals and Agent Optimizer runs that return scored code diffs
- Data Workshop with dataset preparation and quality checks
- Model training with recommended settings for LoRA or full fine-tuning
- Benchmarks against your current model on the same eval set
- Hosted OpenAI-compatible inference and downloadable model weights
- Console, CLI and MCP server access
- OpenAI-compatible API for trained models
| - Record every model call, tool use and agent step
- Replay decisions from original context
- Policy enforcement before tool execution
- Turn failures into regression tests
- Tamper-evident audit records
- MCP gateway integration
- Incident tracking from failed runs
- Prompt registry and playground
- Fleet control room and policy simulation
- Signed audit records with integrity verification
|
|---|
| What makes it different | - Own your trained models with downloadable weights and checkpoints
- Zero third-party training with data never leaving your pipeline
- Open source and self-hostable with Docker Compose
- Built by security and defense intelligence pioneers
| - Replay over retell: hold context fixed, run the model fresh to verify behavior actually changed
- Block before it ships: enforce policy at release time, not after a failure in production
- Self-hosted by default: full control with deployment in customer's own network
- Tamper-evident records checkable without Runback
- Turn production failures into future test cases
|
| Community votes | 0 | 0 |
| DR (Domain Rating by Ahrefs) | 44 ● no change since last check | Not checked yet ● no change since last check |
| Trust Flow (Majestic) | 10 | 0 |
| Citation Flow (Majestic) | 35 | 9 |
| Referring domains (Majestic) | 201 | 1 |