| What it does | The continual learning engine for AI agents
| Identity and flight recorder for AI agents
|
|---|
| Stage | Not given
| MVP
|
|---|
| Founded | Not given
| Not given
|
|---|
| Based in | Not given
| US
|
|---|
| Pricing model | Freemium
| Freemium
|
|---|
| Pricing | RELAI Free: $0. RELAI Enterprise: Custom pricing.
| Free: up to 5 agents, no card required. Starter: $49 USD per month for up to 25 agents. Pro: $199 USD per month for unlimited agents. Per-agent pricing with no charge per request or token.
|
|---|
| Key features | - Capture signals from code, requirements, and production traces
- Build and maintain EvalsWiki as versioned behavioral knowledge
- Find missing eval coverage and unclear expectations
- Generate complete evals with scenarios, environments, and evaluators
- Optimize harness across prompts, tools, workflows, and code
- Multi-objective optimization for quality, cost, latency, and KPIs
- Model selection, migration, and routing with evaluation
- Regression-controlled improvement with evaluation evidence
- Versioned storage in repository with pull requests
| - Stable signed identity per agent tied to organization
- Complete event record: runs, model calls, tool calls, errors, retries with timing
- Events hashed, chained and signed before leaving host
- Real-time fleet view with error rates and instance comparison
- Non-proxy architecture: agents call models directly
- SDK or OpenTelemetry exporter integration
- Revoke agent credentials with immediate partner visibility
- Suspend agent instance and refuse further events
- Signed evidence packages for audits and auditor review
- Organization data export API for compliance
|
|---|
| What makes it different | - Turns failures into replayable learning environments
- Prevents regressions by keeping previous evals in the loop
- Integrates with existing observability stacks and traces
- Preserves business intelligence versioned in repository
- Adapts harness to model changes without redefining use cases
| - Non-proxy design means agents keep running even if Entruva is down
- Per-agent pricing, not per request or token, so more recording costs nothing more
- Records are signed at collection and exportable without operator cooperation
- Telemetry predicts task failure within a domain better than model alone
- Setup takes five minutes with SDK or OpenTelemetry exporter
|
|---|
| Free plan | Not given
| Not given
|
|---|
| Open source | Not given
| Not given
|
|---|
| Platforms | Not given
| Not given
|
|---|
| Public API | Not given
| Not given
|
| Community votes | 0 | 0 |
| DR (Domain Rating by Ahrefs) | 22 ▼ −1 (down) since last check | Not checked yet ● no change since last check |
| Trust Flow (Majestic) | 13 | 0 |
| Citation Flow (Majestic) | 25 | 0 |
| Referring domains (Majestic) | 79 | 0 |