The full-trace workflow engine for AI agents

The graph decides. The model writes. The trace remembers.

Fulltrace runs AI agents inside task graphs whose shape is fixed in code. A model writes the content of each step. Code decides what runs next, what passes, and what may be applied. One authority object grants a change, one append-only ledger records it, and every node, stage and tool call lands in a trace you can open afterwards and replay.

The demo

Fulltrace Studio, toured live.

One continuous walk through Fulltrace Studio: both modes, every tab, a workflow, a run, and the findings it hands back. Every screen is a real capture of the packaged extension, and the cursor goes exactly where the capture pressed. It plays full screen, where the arrow keys step through and the space bar pauses.

Watch the demo

Opens full screen. Escape closes it.

Demo available on desktop or tablet only.

Fulltrace Studio
Fulltrace Studio in the code editor: the sidebar with the system's health and usage limits, and Studio open beside it on Operate > Work, the workflow catalogue.
0:00

How it works

Context. Runs. Changes.

Three questions, one record. Fulltrace keeps what your agents knew, what they did, what they were allowed to change, and what every step cost.

01

What they know

Your project context is served to every AI assistant you use from one place, with shared memory, commands and hooks beside it. Every read is counted, so you can see which files your agents use and which they ignore.

02

What they do

Work runs as a task graph fixed in code, with a model chosen per stage and a price estimated up front. Launch it by hand, or arm it on a schedule inside spend caps you set. Every node, stage and tool call lands as a span on the run's real clock.

03

What they change

A proposed change is frozen as a hashed artefact, simulated through the project's own build with a mandatory revert, and applied only under a grant bound to that artefact. Nothing lands without a person.

The factory

Specs are promises. Traces are receipts.

Spec-driven development writes the intent down, hands it to an agent and reviews what comes back. Fulltrace picks up where that stops: the checks run while the work does. The product itself is built the same way. Every slice of Fulltrace ships through kickoff, decision gates, adversarial review and a dated closeout, and the record is public.