Codag

Codag provides drop-in log compression for AI agents by converting extensive infrastructure logs into a schema-valid, incident-focused capsule. It aims to keep only the meaningful lines and tag them by role so agents can reason over structured data rather than raw logs.

Codag compresses up to 1.2 million log lines into a schema-valid IncidentCapsule containing about 3,300 tokens in under a second. Each kept line is tagged with a role (root_cause, trigger, consequence) and linked to the real log line number, preserving causality. It works as a drop-in, wrapping existing log streams over HTTPS or via MCP/Claude integration, and returns a structured capsule that can be consumed by any agent or LLM without altering the agent configuration. Supports various line-oriented formats (JSON logs, syslog, Hadoop/Spark/HDFS, Kubernetes, etc.).

Who it’s for: Developers and teams building AI-assisted incident response and debugging workflows who need compact, structured log capsules for agents.

mentions of integration capabilities, public product features, and a free offering; no explicit funding or hiring data provided