Fluiq

The product is a control plane for AI agents that provides observability, security scanning, and evaluation for LLM calls. It traces tokens, cost and latency per agent node, blocks prompt injections, jailbreaks and PII/secret leaks, and scores outputs against custom criteria. It is for developers and engineering teams building agent applications with frameworks such as LangGraph, CrewAI and MCP. It is delivered as a SaaS platform and infrastructure API integrated with two lines of Python and 14 integrations with no wrappers.

Key features

• Trace every LLM call cost latency
• Per-node token attribution
• p50 p95 p99 latency tracking
• Multi-agent DAG visualization
• Real-time trace streaming
• Block prompt injection and jailbreaks
• Redact PII and secrets
• Scan RAG poisoning and indirect injection
• Detect tool misuse and exfiltration
• Enforce tool allowlist
• Detect cross-agent trust escalation
• Scan images and media
• Tunable guardrail policies and thresholds
• Fails open on scanner error
• Score responses and agent runs
• Evaluation datasets for agent runs
• Prompt template versioning and deployment
• Push eval and security alerts to Slack