Skills

Research Toolkit Skills

Choose a research skill, send a short input, and the server-side runtime will return a structured result. This page is the product shell; model routing can be attached behind the same API later.

Back to Research Toolkit

Runtime

Available skills

Connecting to the runtime...

Status: checking Runs: --

Run

Choose a skill

Select a skill from the list, then send a short research idea.

Running a skill currently calls the Research Toolkit backend. It does not yet call an external language model; the next layer is a model adapter that sends the skill output into a chosen LLM provider. 

No run yet.

Access

Invite-only testing

Sign in with an email and invite code before using the model-adapter test path.

Auth: checking Ask limit: --

Ask With Skill

Test the model adapter path

This test first runs the selected skill, then sends the skill-generated context to a provider adapter. The default provider is a local mock, so it proves the flow without spending API credits. 

No model-adapter run yet.

Stage 1

Front door only

  • Keep the navigation page on the same domain.
  • Reserve the path for later skill execution.
  • Use the current server for the first rollout.

Stage 2

Attach runtime

  • Reverse proxy the same path to a local service.
  • Add skill execution, logs, and usage tracking.
  • Keep the public URL unchanged.

Stage 3

Split only if needed

  • Move heavy PDF or OCR jobs off the main box.
  • Separate vector indexing if it starts to churn.
  • Add capacity only when the workload demands it.