SafeForge AI
SafeForge AI is an agent-first system safety platform for safety-critical industries such as rail, defence, aviation, and oil & gas. It combines hazard management, live bow-tie risk analysis, risk workshops, and assurance workflows with an AI agent that reads the hazard log through governed tools. According to its homepage, the agent can draft summaries and change requests, follow the conversation on screen, and stage AI suggestions as unconfirmed until a human reviewer accepts them. SafeForge emphasizes human-reviewed, defensible change control. Every AI-drafted change request is raised by the agent but accepted by a person, and the platform supports governed checks against good international practice rather than issuing verdicts. It also offers connectors for DOORS and SharePoint, allowing teams to work with existing engineering and document management systems. The product is available as a beta release with a 14-day trial and freemium pricing. The tool supports MCP, so users can work inside SafeForge or in an AI assistant their team already uses. Example use includes adding missed causes to a key-risk bow-tie based on published investigations, then staging the change as AI-unconfirmed for team review.
- Category
- Product safety evaluation
- Pricing
- Freemium
Problem solved
Safety-critical engineering teams need to maintain hazard logs, bow-tie risk models, and change records that are traceable, auditable, and defensible. Manual safety analysis and documentation can be slow and error-prone, while ungoverned AI suggestions risk introducing unsupported changes. SafeForge solves this by providing an agent-first hazard management platform where AI drafts summaries and change requests, but every change remains human-reviewed and connected to existing tools like DOORS and SharePoint.
Use cases
- Maintain a live hazard log for rail signalling systems and other safety-critical assets
- Generate draft hazard summaries and change requests from an existing hazard log
- Conduct bow-tie risk analysis for scenarios such as Signal Passed at Danger (SPAD) and derailment
- Run risk workshops with staged AI-suggested controls that require team review
- Integrate hazard management with DOORS requirements data and SharePoint documentation
- Support assurance and audit trails for safety-critical projects
- Use MCP-enabled AI assistants to query governed hazard data and propose changes
Audience
- Enterprise
- Business Users
Pros
- Agent-first workflow drafts hazard summaries and change requests while keeping a human reviewer in the loop.
- Live bow-tie risk analysis with support for real-world incident examples and staged AI suggestions.
- Designed for safety-critical industries including rail, defence, aviation, and oil & gas.
- Connectors for DOORS and SharePoint plus MCP support for existing AI assistants.
- Governed checks and change control aimed at defensible, auditable safety records.
Cons
- Currently in beta, so features and stability may evolve.
- No public GitHub repository or open-source code was referenced on the homepage.
- AI suggestions are intentionally not auto-accepted; all changes require human review.
- Integrations listed are limited to specific tools such as DOORS and SharePoint.
- Pricing is freemium, but public details beyond a trial and beta discount are limited.
Official website: https://safeforge.ai/
Canonical LambdaBase page: https://www.lambdabase.com/tools/safeforge-ai