JamJet Cloud

The cockpit
for agent runs.

Trace, govern, approve, replay, and audit agent runs across your team. The open-source runtime runs your agents. Cloud shows what happened, what was blocked, what needs approval, where cost is going, and what can be replayed.

pip install jamjet[openai] · two lines, then the dashboard fills in.

OSS runtime stays free. Cloud is the cockpit.

Two lines

Then the dashboard fills in.

import jamjet.cloud as jamjet
from openai import OpenAI

jamjet.configure(api_key="jj_xxxx", project="my-agent")  # 1
# 2: every OpenAI / Anthropic call is captured automatically

OpenAI().chat.completions.create(
    model="gpt-4o-mini",
    messages=[{"role": "user", "content": "hello"}],
)

Open app.jamjet.dev/dashboard/traces. The call appears within ~5 seconds with model, token counts, cost, latency, and the policy decisions that ran.

In Cloud today

Six surfaces for the team that ships agents.

Trace timeline

Every model call, tool call, policy decision, approval, and checkpoint — in order, with cost and latency on each row.

Policy violations

What was blocked, why, and how to fix it. 4-level rule hierarchy with diff-friendly history.

Approval queue

High-risk actions wait. Reviewer assignment, SLA, and per-tool routing. Decisions land in the audit log.

Cost guardrails

Track and enforce spend per agent, run, and environment. Alerts before limits hit; hard caps stop the call.

Audit export

Evidence packages for incidents, reviews, and compliance. CSV, JSON, PDF, or SIEM-ready stream.

Hosted Engram

Shared memory for agents and teams. Same MCP API, no Postgres to run, retention controls per project.

OSS vs Cloud

Everything local stays free.

The runtime, Engram, and SDKs are Apache 2.0 forever. Cloud adds the team surface — shared traces, hosted memory, retained audit, and approvals across people.

Capability

OSS runtime

JamJet Cloud

Local execution & durability

Yes

Policy engine

Yes

Audit log (local)

Yes

Engram memory

Self-hosted

Hosted

Trace dashboard

Local web companion

Team-wide

Approval queue across reviewers

API only

Hosted UI + SLA

Retention & SIEM export

Manual

Built-in

SSO / RBAC / projects

—

Team / Enterprise

Cross-agent trace propagation

Manual

W3C traceparent

Plans

Start free. Pay when your team needs it.

OSS

Free, forever

Apache 2.0. Self-host the runtime and Engram. Unlimited.

Runtime, Engram, all SDKs
Local traces & audit
Policy + approvals via API
Community Discord

View on GitHub

Cloud Free

For solo developers. No card, no commitment.

1k traces / month
7-day retention
Hosted Engram (1 project)
Approval queue (1 reviewer)

Team

Coming soon

For small teams shipping agents together.

Higher trace volume
Projects + reviewer roles
Retention controls
Cost guardrails per project

Join waitlist

Enterprise

Talk to us

For regulated teams & self-hosted control planes.

SSO / SAML / OIDC
RBAC + tenant isolation
SIEM export, audit retention
Self-hosted option

Roadmap

What's coming next.

Multi-agent network graph (Q3 2026). Force-directed view of how your agents communicate — nodes are agents and MCP servers, edges show calls, costs, error rates. Drill down to traces.
Java cloud SDK (Q3 2026). Same drop-in for Spring AI / LangChain4j. Auto-instruments ChatClient and ChatModel.
Cross-agent trace propagation. When agent A calls agent B (over A2A or MCP), traces link automatically. W3C traceparent + custom tracestate.
Centralized policy with delegation chains. Visualize who-authorized-what. Compliance-grade audit.
Trace replay. Re-run any trace with input recordings. Debug incidents or test prompt changes against real history.
OTel GenAI ingestion. Point existing Phoenix / OpenLLMetry / Langfuse-instrumented apps at JamJet without an SDK migration.

Send your first trace.

Two lines of Python and the dashboard fills in. No credit card.

The cockpitfor agent runs.