Debug AI agents · Capture · Compare · Replay
Wrap any agent in one command.
Capture every event. Compare runs. Replay from any state.
Cryptographically signed end-to-end. SOC 2 Type II + EU AI Act Article 12 ready.
Scroll to continue ↓Debug AI agents · Capture · Compare · Replay
Wrap any agent in one command.
Capture every event. Compare runs. Replay from any state.
Cryptographically signed end-to-end. SOC 2 Type II + EU AI Act Article 12 ready.
Scroll to continue ↓
Cryptographic audit chain. Deterministic replay.
Persistent memory. For any AI agent, anywhere.
Five audience streams below. Each has setup, pricing, and demos specific to you.
Beyond Capture
The CLI capture-and-audit loop is just the first 10% of what SteelSpine does. Underneath it is a five-layer infrastructure stack — every piece runs locally, no cloud dependency, no vendor lock-in.
Wrap any agent or command. Stream stdout/stderr to a hash-chained event log. Replay offline against any captured state.
steelspine run · replay-run · branch-create
HMAC-SHA256 + Ed25519 chain. Tamper-evident. Independently verifiable by an auditor with just the public key. Produces Article 12-ready tamper-evident audit records out of the box. Optional hardening: compliance_mode auto-enables RFC 3161 timestamping via eIDAS-accredited TSA. The integrity layer is classical (HMAC-SHA256 + Ed25519); it is not quantum-resistant. Tamper-evidence plus independent verification protect integrity-after-capture and let a third party verify; because the signing key is held by the operator, this is not non-repudiation against the operator — key custody (HSM / KMS / third-party timestamping) is the control for that.
verify-run · pack-create · pack-verify
Transparent proxy in front of any OpenAI-compatible LLM. Auto-injects relevant context into every prompt. Promotes durable facts to long-term entity store. The same agent remembers across sessions.
memory-agent · memory recall · entities
OpenTelemetry receiver for LangChain, LlamaIndex, CrewAI, Claude Code, and 50+ OTel-instrumented agents. Native MCP server exposes 8 inspection tools to Cursor, Claude Code, and Windsurf — one config, three clients. Plus filesystem-drop, passive-watch, raw-log-capture.
otel-receiver · mcp-server · capture-pipe
Branch from any captured state. Simulate alternate paths. What-if any decision your agent made — explored offline, no live API costs.
branch-create · simulate · replay-branch
No cloud uploads. No telemetry to vendors. Your agent runs, your captures, your memory, your audits — all stay on your machine. Works offline. Works in air-gapped environments. Ships with the bundle.
~/.prime/ · open architecture
Native integrations: Claude Code (OTEL + MCP), Docker (watch), Git/CI (compare --strict), Cursor (MCP + agent-trace), Windsurf (MCP). One MCP server, three IDE/CLI clients. Plus 50+ OTEL-instrumented frameworks (LangChain, LangGraph, CrewAI, LlamaIndex, OpenAI Agents SDK, Haystack, DSPy). · Not yet supported: hosted UIs (ChatGPT.com, Claude.ai web). See integrations for setup guides.
Click a tile. Land on a page built for your situation, with setup, pricing, and demos specific to you.
Cryptographically-signed audit trails for high-risk AI systems. Multi-vendor, local-first, regulator-verifiable. Free — reach out at hello@steelspine.ai.
Wrap any agent. Find where it failed. Replay deterministically. Verify it wasn't tampered with. Free — reach out at hello@steelspine.ai.
NPCs that remember across sessions. Cryptographically-verified anti-cheat. Replayable world state for evolving questlines. Free — reach out at hello@steelspine.ai.
compare --strict in CI exits non-zero on regression. GitHub Actions, GitLab, Jenkins. Block bad agent behavior before merge.
Mem0 / Letta / Zep alternative with local-first storage and cryptographic audit. One URL change. Any OpenAI-compatible LLM.
Claude Code, Docker, Git/CI, Cursor, Windsurf, plus 50+ OpenTelemetry-instrumented frameworks. One MCP server, three IDE/CLI clients.
Not sure which fits? Keep scrolling for the full SteelSpine pitch, or email hello@steelspine.ai.