Footprint. From research to report.
A platform for composing AI-agent pipelines. Each task has its own model and tools; the workflow takes a brief through to an assembled, reviewed and, when needed, translated report.
The problem to solve.
A company report requires different sources, research tasks and analyses. Some tasks can run together; others need earlier results. The workflow needs to change without rewriting the application.
Configure workflows through the interface.
The interface defines tasks, tools and dependencies. Independent tasks run in parallel; dependent tasks wait for earlier outputs. Synthesis combines the results into a draft, which then goes through review for missing information and inconsistencies.
Review is a separate stage from synthesis. The chain of dependencies remains traceable in the report.
Inside the system.
Incoming brief
The request specifies the company, sector, and questions the report needs to address.
Composing the pipeline
The interface defines which tasks run in parallel and which depend on the output of an earlier step.
Agents at work
Each agent has a defined task, model and tools, connected through plugins and MCP (Model Context Protocol). Tasks include business-data research, sector analysis and collecting public signals.
Combining the results
An assembly step combines the outputs into a single draft.
Critical review and translation
A review step checks the draft for missing information and internal inconsistencies. Translation follows when required.
Report delivered
The final document retains a traceable record of the workflow and its dependencies.
What it enables.
Users compose and modify workflows without requesting a code change for every adjustment.
The interface and API use the same workflow model: new steps can be added without rewriting the rest of the pipeline.