Unverified50% confidenceFactExact time
T3 Code's MCP layer allows one Harness to control another Harness, for example using Claude Code to launch a sub-Agent that uses another model.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
10/30/2026
First Seen
Valid until: 1/28/2027
Sources
Is Harness Dead? Why the Creator of T3 Code Says Stop Obsessing Over AI Coding Tools
youtubeTheo - t3․gg10/30/2026
Related Entities
Related Claims
UnverifiedT3 Code does not add new tools to the Harness, but instead lets the Agent access data and systems outside the Harness via MCP, such as Supabase and PostHog.78% similarUnverifiedT3 Code's orchestration layer is implemented as a flexible MCP, allowing agents to communicate with each other, manage other threads, and be controlled remotely via external tools (such as an iMessage bot).76% similarUnverifiedBefore MCP appeared, if developers wanted models such as Claude and ChatGPT to call third-party services, they usually needed to write plugin or function-calling logic separately for each model platform.69% similarUnverifiedClaude Code提供Skills功能作为MCP的替代方案，提供类似能力但不会占用整个上下文窗口66% similarUnverifiedMCP services mainly extend the tool boundaries of the Agent's execution layer, while scheduled tasks and trigger conditions belong to the scheduling logic of the orchestration layer.65% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1118770API
curl https://kongchang.com/api/v1/knowledge/claims/1118770MCP
get_claim(id=1118770)