Camunda announces ProcessOS, an AI-first business transformation operating system
Agentic orchestration platform provider announced ProcessOS, a new intelligence layer that the company says is leading to a “massive redesign” of processes and associated workflows.
According to the company, beyond just AI job assistance, companies will have to rethink what tasks should be done, and whether agents or humans should do them. “Simply incorporating AI into legacy processes entails technical and organizational debt, adding complexity, vulnerability, and cost. AI-first process transformation must work backward from results, not forward from what is happening now,” Camunda wrote in his announcement.
Designed for enterprise deployments, ProcessOS ensures full governance and control:
- Design verification: visual process models that clearly show what steps AI performs, under what conditions, and which steps involve humans
- Someone in the loop for any change: All process changes are reviewed and approved by people before they reach the product
- Recycle and learn built-in: ProcessOS prioritizes patterns of authorized processes and interfaces, evolving with each human feedback loop
ProcessOS runs natively on AWS, with deep integration into Amazon Bedrock and Amazon Bedrock AgentCore for underlying models, agent memory, identity, and gateway services. Camunda is available on the AWS Marketplace with production-ready references for Amazon EKS, ECS, and EC2.
ProcessOS is available in closed beta for select businesses starting today. Organizations can register their interest to participate at camunda.com/process-os.
Hadrian Releases OpenHack, Democratizing AI Vulnerability Discovery
Hadrian released OpenHack, an open source AI-powered source code review tool that works within Claude Code, Codex, and Cursor.
The temptation when giving a solid LLM a codebase is to allow it to evolve. “Read this repo and tell me what’s at stake.” It will reveal something. The output will be a mixture of audible bugs, forgotten bugs, mis-explained real bugs, and the occasional sharp insight. Experimenting takes longer than learning the code yourself.
We found two failure modes driving most of that noise:
-
Unchecked commands. The agent does not know which question to answer, so it answers all with low confidence.
-
Self-discovery. The same agent that proposed the bug decides whether the bug is real. There is no independent check.
OpenHack’s workflow is designed to fix those two things. The review is the first case: each unit of the agent’s work is one routing unit, one expert, and one proof question. And the agent proposing the acquisition is not the agent agreeing.
How OpenHack works:
-
Context-based scope. Every unit of work is one routing unit, one expert, and one proof question. There are no hidden hints asking the model to find anything wrong.
-
Independent decision. The agent proposing the acquisition is not the agent agreeing. A separate triage agent reviews each candidate against the original evidence before a finding is recorded.
-
A testable artifact route. Outputs, status backlogs, expert results, decisions, and findings all live as blank files on disk. The full review can be studied end to end.
-
Harness-agnostic, model-agnostic. Works within Codex Claude, Codex, or Cursor, or any model supported by the harness.
OpenHack is available immediately at github.com/hadiriansecurity/
Devart releases dbForge 2026.1
dbForge’s AI-powered skills
-
Natural language conversion to SQL
-
Context-aware SQL query execution
-
Developing pre-written SQL queries
-
Finding unwanted or missing references
-
Detailed SQL information
-
Error analysis and troubleshooting
-
Integrated AI discussion of database and SQL related queries.
dbForge 2026.1 is available today with dbForge Edge, dbForge
