Integrations

Attaches to the stack you already run.

LonsLabs reads traces, opens reviews, and enforces permissions through the tools your team uses today.

Agent frameworks

LangGraph, LlamaIndex, CrewAI, AutoGen, the OpenAI Agents SDK, Anthropic's Agent SDK, and custom orchestrators. LonsLabs reads traces from the framework you already run; it does not require a migration.

Model providers

Anthropic, OpenAI, Google, Mistral, and open-weight models served through vLLM or Bedrock. The loop is model-agnostic — a lesson applies to the prompt or policy, whichever model executes it.

Observability & tracing

OpenTelemetry, Langfuse, LangSmith, Arize Phoenix, Datadog. Existing traces become LonsLabs evidence; you do not instrument twice.

Ticketing & feedback

Zendesk, Intercom, Jira, Linear, GitHub Issues. A ticket that flags a failure is linked to the resulting change so the loop is visible from the ticket.

Version control & review

GitHub, GitLab, Bitbucket. Artifact edits open as pull or merge requests with the evidence in the description; approvals map to your existing CODEOWNERS.

Identity & permissions

Okta, Azure AD, Google Workspace via SAML or OIDC. Artifact ownership and approval rules follow the groups you already maintain.

Evaluation & CI

GitHub Actions, GitLab CI, CircleCI, Buildkite. The harness runs as a required check so unverified changes cannot merge.

Messaging & alerts

Slack and Microsoft Teams. Owners are notified where they work; approvals can be granted from the thread.

Deployment

Cloud, private cloud, or on-premises.

Most teams start on LonsLabs Cloud, hosted in the EU. Regulated customers can run the control plane in their own VPC on AWS, Azure, or GCP, or fully on-premises. In every option, customer content stays in the region you choose.

Trace ingestion through OpenTelemetry or native SDKs (Python, TypeScript).
Artifact edits delivered as pull requests to your repositories.
Harness execution in your CI or in an isolated LonsLabs runner.
Audit export to your SIEM.

python · lonslabs sdkv1
1from lonslabs import observe
2
3@observe(agent="refund-agent")
4def handle(ticket):
5    return agent.run(ticket)
6
7# failures flagged by humans or monitors
8# become harness cases automatically

Not listed?

We add integrations with customers.

If your stack is not on this page, tell us what you run. Most connectors take days, not quarters.

Talk to us