Knows the trade language.
Won't confuse a tangent for a guy. Knows what makeready means. Knows what midspan height implies for the connection it's attached to. Doesn't need to be prompted into the domain.
Agent
PreviewFieldTrip Agent is a purpose-built agent that knows the platform from the inside out. Same MCP server. Same workspace token. Same audit log. The difference is the system prompt, the tool selection, and the temperament — all tuned for OSP workflows rather than generic chat.
A general-purpose assistant connected to FieldTrip via MCP can do a lot. A purpose-built agent does the same things, faster, with fewer wrong turns, and without you teaching it the domain every time.
Won't confuse a tangent for a guy. Knows what makeready means. Knows what midspan height implies for the connection it's attached to. Doesn't need to be prompted into the domain.
Understands pipeline stages, signoff gates, and confidence thresholds. Won't try to writeback before signoff. Won't suggest skipping a review on a low-confidence frame. Acts like a teammate, not a tourist.
Runs against the same MCP server, same workspace token, same audit log. Every action it takes shows up where your QC leads can see it.
Built on Google's Agent Development Kit in Go, running on Gemini 2.5 Flash. Tuned for fast, audit-able decisions on real OSP workflows — not generic chat.
You have three ways to drive FieldTrip programmatically. MCP is for connecting your own assistant (Claude, ChatGPT, Cursor). CLI is for humans and scripts that want a terminal surface. The Agent is for teams that want a FieldTrip-native experience without assembling the prompt themselves.
All three speak the same protocol, hit the same audit log, and carry the same workspace scopes. Pick the one that fits how your team works — or use all three.
We're onboarding a small number of engineering firms onto the agent first — to validate workflows, tune the temperament, and harden the production wiring. Tell us what your team does and we'll see if it's a fit.