Home / Integrations

Works with the stack your team already uses

SteelSpine integrates natively with the four tools defining modern AI engineering in 2026, plus 50+ OpenTelemetry-instrumented agent frameworks. Native receivers, drop-in adapters, MCP server. No SDK lock-in.

The 4 integrations every modern AI team uses

Claude Code

Native · MCP + OpenTelemetry

Anthropic's CLI agent. Two integration paths: enable Claude Code's native OpenTelemetry signals and point them at SteelSpine's receiver, or configure SteelSpine's MCP server so Claude Code can directly query your audit history. Both work out of the box.

Setup guide →

Docker

Native · prime-watch + OTEL

steelspine watch passively monitors Docker containers, log files, systemd services, and process trees. Agents running inside containers can also emit OTLP to SteelSpine running outside. The audit chain spans the container boundary cleanly.

OTEL setup →

Git / CI

Native · compare --strict

steelspine compare --strict exits non-zero when a new run is a regression versus baseline. Drop into GitHub Actions, GitLab CI, or Jenkins to gate PRs on agent regression. Audit evidence travels with the pull request.

CI/CD recipe →

Cursor

MCP server + agent-trace adapter

Two adapters: the SteelSpine MCP server exposes audit and inspection tools so Cursor's AI assistant can directly query your captured runs. The agent-trace adapter ingests Cursor's AI vs human code attribution records into the signed audit chain. Both work in the same project.

Setup guide →

Plus 50+ OpenTelemetry-instrumented frameworks

Any agent framework that emits OpenTelemetry traces over OTLP/HTTP can funnel into SteelSpine with zero code changes. Just point OTEL_EXPORTER_OTLP_ENDPOINT at the receiver.

Verified frameworks (2026):

Generic OTEL setup steps:

# 1. Start the SteelSpine OTEL receiver
steelspine otel-receiver --port 4318 --project my-agent

# 2. Configure your agent to emit to SteelSpine
export OTEL_EXPORTER_OTLP_ENDPOINT=http://localhost:4318
export OTEL_EXPORTER_OTLP_PROTOCOL=http/json

# 3. Run your agent as normal
python3 my_langchain_agent.py

See the full OpenTelemetry guide for framework-specific configuration and troubleshooting.

What about hosted UIs?

SteelSpine does not currently integrate with hosted AI UIs (ChatGPT.com, Claude.ai web). Those products do not expose runtime telemetry to external observability tools.

For workflows mixing hosted UI and local agents, the integration pattern is: do the agent runtime locally (Claude Code, Cursor with MCP, LangChain agents), capture with SteelSpine, use hosted UIs as conversational interfaces. The audit chain lives where the work happens.

If your stack uses something not listed here: if your framework emits OpenTelemetry, it works with SteelSpine. If it doesn't, the prime-emit HTTP API accepts arbitrary structured events. Ask us in hello@steelspine.ai and we'll confirm.