Platform
One execution record. Three connected capabilities.
JeraViz is built around a single record of what your agents did. Trace is the first product. Guard and Eval will use the same projects, keys, telemetry and workspace, and are labelled as roadmap until they exist.
Flagship product · Launch scope
JeraViz Trace
The flagship product and the first release. It captures every execution as a trace and gives engineers the views to understand it.
- Distributed agent execution tracing
- End-to-end execution timelines
- Tool-call and model-request inspection
- Token, cost and latency analytics
- Execution history and search
- Agent-to-agent visibility (graph view planned)
- ticket-resolution
- triage-agent
- classify-ticket
- retrieval-agent
- vector_search
- kb_lookup
- synthesize-context
- refund-agent
- plan-refund
- orders_api.get_order
- payments_api.create_refund
- compose-reply
Connected capability · Roadmap
JeraViz Guard
Planned. Governs what agents are allowed to do, using the same workspace and telemetry as Trace. Not available and not designed in detail yet.
- Agent permissions and access policies
- Tool execution controls
- Human approval workflows
- Risk-based execution policies
- Policy violation alerts and audit trails
Not available · direction only
Guard would evaluate each tool call against policies and record the decision as an event on the same trace, so the audit trail and the execution record are one dataset.
No screens are shown because none have been built. Scope, scoring methods and release timing are not decided.
Get notified about early accessConnected capability · Roadmap
JeraViz Eval
Planned. Turns recorded executions into test cases and measures quality over time. Direction only: scoring methods have not been chosen.
- Scenario-based testing and evaluation datasets
- Execution quality scoring
- Regression testing before deployment
- Benchmark comparisons and reliability reports
Not available · direction only
Eval would let you save a recorded execution as a test case, run a dataset against two agent versions and compare results, using the same traces you already inspect.
No screens are shown because none have been built. Scope, scoring methods and release timing are not decided.
Get notified about early accessData model
What Trace captures.
Everything is a span in a trace; an execution is one run. Choose how much text is stored with a capture mode.
| Record | Fields | Note |
|---|---|---|
| Execution | name, status, start and end, duration, tags, optional session id | All plans |
| Span | kind, name, parent, start, duration, status, attributes | All plans |
| Model request | model, provider, input and output tokens, finish reason, optional prompt and response text | Text depends on capture mode |
| Tool call | tool name, optional arguments and result, retries, error category | Payload depends on capture mode |
| Event | timestamp, type (exception, retry, log), attributes | All plans |
Capture modes (per project)
- Full
- Stores prompts, responses, tool arguments and results (subject to size limits and redaction rules).
- Metadata only
- Stores timing, status, tokens and names, but no payload text. Recommended for sensitive workloads.
- Hashed payloads
- Stores a hash of each payload so identical inputs can be recognized without storing the text.
Designed for isolation, auditability and scale.
The proposed data flow and the principles behind it. Statuses are explicit: these are design principles for the first release, not audited controls.
- 01AI ApplicationSDK or OpenTelemetry exporter
- 02Telemetry IngestionAuthenticated, validated, rate-limited
- 03Event ProcessingNormalize, price tokens, apply privacy rules
- 04Trace StorageColumnar event store + object storage
- 05AnalyticsRollups, percentiles, cost
- 06JeraViz DashboardExplorer, timeline, analytics
Secure data transmission
Launch scopeTLS for public endpoints; project-scoped API keys on ingestion.
By design
Workspace isolation
Launch scopeEvery record is scoped to a workspace; enforced in application code and database policy, and tested.
By design
Access controls
Launch scopeRole-based access at workspace level (owner, admin, developer, viewer).
By design
Encryption
Launch scopeEncryption at rest with managed keys for databases, object storage and queues.
By design
Auditability
PlannedAdministrative actions are recorded from the start; a user-facing audit log follows.
Recording at launch, UI planned
Data retention
Launch scopeRetention set by plan; deletion on request; per-project capture modes limit what is stored.
Per-project retention planned
Release status
What is planned, in order.
No dates are promised. Order reflects dependencies: later phases build on the data captured by earlier ones.
Phase 1
Foundation
Launch scope- Authentication, workspaces, projects, API keys
- Event ingestion (SDK and OTLP)
- Trace explorer and execution details
- Core analytics: volume, latency, tokens, cost
Phase 2
Observability
Planned- Advanced filtering and saved views
- Agent execution graph
- Tool-call, model-request and error views
- Alerts, SDK and framework expansion
Phase 3
Intelligence
Roadmap- Advanced analytics and comparisons
- Evaluation datasets and runs
- Regression testing
- Agent performance comparisons
Phase 4
Governance
Roadmap- Permission controls and policy engine
- Human approval workflows
- Enterprise audit capabilities
Build agents with visibility. Operate them with confidence.
Understand what your agents do, identify what needs improvement and build more reliable AI systems with JeraViz.
JeraViz Trace is in development. Sign-up is not open yet; the early-access page explains where things stand.