Graylog vs Panther

About Panther

Panther is an AI SOC platform that ingests and normalizes logs at petabyte scale, lets teams write detection rules in Python, and automatically creates detections using AI and natural language. Panther is a cloud-native SIEM featuring detections-as-code and advanced search, with natively supported log sources and alert destinations. Panther ingests terabytes of security events per day into a structured data lake to power real-time detections and investigations. Panther uses a security data lake architecture that provides long-term retention. Customers own their data and can query it at any time, and retention policies in Panther are configurable to meet compliance requirements.^[2][3]

Panther says that cloud security logs, EDR portals, and MDR consoles each exist in silos with no central view. Panther states that analysts manually piece context together, often missing critical details. Panther describes an agentic SOC in which agents hunt the full data lake on a regular cadence, surfacing findings that flow into new detections through a governed workflow. In Panther, every alert is investigated with full context, and confirmed benign alerts close automatically. When a finding escalates in Panther, the agent assembles the full investigation across multiple systems, and analysts review and guide agentic runbooks to take action.^[6]

Panther's AI SOC Agent runs on a schedule, responds to natural language queries, and takes action with complete context. Every investigation by the AI SOC Agent pulls from the customer's data lake, detection logic, and connected tools via MCP. When an alert fires, Panther automatically gathers evidence across the data lake, detection logic, alert history, and connected tools. Panther AI queries the data lake, reviews detection logic, and pulls enrichment from connected tools, delivering a risk classification with reasoning. Any analyst can investigate through a natural language interface in Panther without writing SQL, PantherFlow, or any other query syntax. Panther exposes its own MCP server, so external agents can query alerts, detections, and log data directly. When the risk classification of the AI SOC Agent falls below a threshold the customer configures, the alert closes automatically.^[6]

Panther generates new detection rules from natural language, identifies recurring false positives, and submits tuning changes as reviewable Python code through the customer's existing GitHub workflow. The Panther agent traces an issue back to the specific Python rule responsible and proposes an edit as reviewable code through the existing GitHub workflow, with a diff, explanation, and unit tests. In Panther, nothing deploys without human review. Teams with detection engineers use Detection-as-Code in Panther, with Python rules, version control, and CI/CD workflows. Teams without detection engineers can start with Panther's built-in detection library and AI Detection Builder, then grow into custom rules over time.^[6]

Panther connects to cloud platforms (AWS, GCP, Azure), SaaS tools, endpoints, and network devices. Panther routes alerts to Slack, Jira, PagerDuty, and more. For orchestration and response, Panther provides agentic workflows and runbooks or integrates with external SOAR platforms. Panther offers Okta log monitoring for complete visibility into IdP activity. Panther detections can be enriched with real-time IoCs from Google Threat Intelligence. Panther AI can send messages and search a Slack workspace directly, manage incidents and on-call schedules directly, and search and update a Notion workspace directly. Panther-managed rules for Orca Security are in the panther-analysis GitHub repository. Panther includes built-in dashboards and MITRE ATT&CK mapping, from alert trends to program maturity. In Panther, native access to detection logic means every triage outcome feeds back into the rules that fire.^[6][9][10]

Panther offers two deployment models on one platform, and the choice is where the customer's data lives. In the first model, Panther runs inside the customer's AWS account, against Snowflake or Databricks, and the security data stays in the customer's warehouse. The detection engine, workflows, and agents in Panther operate against that data in place. In the managed model, Panther manages the service and ingests customer data into a dedicated, isolated environment.

Databricks has acquired Panther to accelerate the security lakehouse vision. Panther's AI SOC workflows, 100+ integrations, and detection-as-code capabilities are being integrated into Databricks Lakewatch. Panther requires human approval for every write, and every AI action is logged and reviewable.^[1][6][9]