TASKING's toolchain covers the compiler, debugger, and test harness for safety-critical embedded targets: ISO 26262 automotive, DO-178C avionics, IEC 62443 industrial. These are environments where V&V is not optional, and where every change to the toolchain requires audit evidence. Co-CEO Christoph Herzog confirmed TASKING is now using the Model Context Protocol to expose the toolchain to external AI agents, letting LLMs direct V&V tasks without requiring engineers to script custom integrations for each workflow step.
The mechanism matters because MCP is not a proprietary hook. An AI agent that speaks MCP can now call TASKING's compile, analyze, and test steps the same way it calls any other MCP server. The constraint this removes is real: safety-critical V&V has been resistant to automation not because the tasks are intellectually hard, but because the toolchains are locked and the standard (MISRA, CERT C, AUTOSAR) compliance checks have required human review of every flag. An MCP-connected agent can run a compliance scan, interpret the output, and queue the remediation (the loop that previously consumed a junior engineer).
The risk is that safety standards require traceability back to qualified tools and documented test procedures. A V&V step driven by an LLM agent introduces a new qualification burden. TASKING is presumably addressing this through tool qualification certificates that cover the agent interface, but that is the open question. If they solve the qualification problem cleanly, every DO-178C contractor running TASKING has a path to cut V&V cycle time in half within one tool version cycle.