Iduno

Search Tools

Search Iduno's catalog of privacy-first micro-tools that run entirely in the browser (no data leaves the user's device). Returns matching tools with a run URL the user can open.

Get Tool

Fetch one Iduno tool by its id (UUID) or slug. Returns a run URL AND `inputs` — the exact parameter names this tool reads. Pass those names as run_tool's `inputs` (don't guess). Empty `inputs` = the tool runs interactively (no prefillable params).

Run Tool

Get a ready-to-run link for an Iduno tool with the given inputs prefilled. The user opens it in their browser to run the tool privately (no data leaves their device). For simple stateless tools, a structured result may also be returned directly.

Get Capabilities

Return Iduno's tool-authoring contract: the exact sandbox rules, the required useToolBridge input/output contract, and the FULL list of allowed packages with their canonical import lines. CALL THIS FIRST before writing code for submit_generated_code — it prevents render-fails from a guessed contract or a disallowed/mis-imported package. No arguments; public and read-only.

Create Tool

Build an Iduno micro-tool where IDUNO'S EU PIPELINE writes the code from a spec. IMPORTANT — this is the FALLBACK path: if you can author React/TSX yourself and the tool is pure browser-compute (calculations, file parsing e.g. CSV/SIE, formatting, deterministic logic), DO NOT use create_tool — use submit_generated_code and write the code directly (faster, no ~1–2 min wait, higher reliability, you control quality). Use create_tool ONLY when the tool genuinely needs in-tool AI analysis at runtime

Check Tool Build

Check the build status of a tool created via create_tool (owner-scoped). Returns status ('building'|'done'|'render-failed'|'failed') and, when ready, a run link.

Review Tool

Fetch a tool's spec + build status so you (the user's AI) can review it against the requirements and propose concrete improvements. Read-only — does not change anything. Owner-scoped. Feed your critique back via refine_tool.

Refine Tool

Refine an existing Iduno tool: submit concrete feedback (e.g. from review_tool) and Iduno rebuilds it. Owner-scoped. Returns a new requestId + preview link. Note: one free rebuild per tool; the browser preview stays the primary place a human reviews the generated code.

Submit Generated Code

PREFERRED build path when you (the user's AI) can author React/TSX AND the tool is pure browser-compute. CALL get_capabilities FIRST for the authoritative sandbox contract + the full allowed-package list with exact import lines (prevents render-fails). YOU write the complete tool code and Iduno validates it through its gate-chain (transpile + import/export + render-check in the real privacy sandbox) and hosts it — no Iduno LLM runs, no ~1–2 min pipeline wait, higher reliability. Use create_tool

Check Code

Dry-run a candidate Tool.tsx BEFORE calling submit_generated_code: runs the exact same gate-chain (compile + import/export validation + a REAL render check in the privacy sandbox) but creates or changes NOTHING — no tool is created or updated, no quota/rate-limit shared with submit_generated_code is consumed. Use this to iterate — generate, check, read the real error, fix, check again — as many times as needed, THEN call submit_generated_code once you're confident it renders. CALL get_capabiliti

Submit Tool Order

PREFERRED build path for a calculator/form-style tool (inputs → a formula → a result) that is purely numeric/deterministic (no file parsing, no charts, no multi-view UI, no in-tool AI). CALL get_capabilities FIRST for the exact IR contract (JSON shape, expression grammar, units, known statutory-constant paths) — this tool REJECTS anything that doesn't match it. Instead of writing React, you write a typed ORDER (inputs/compute/outputs as small expressions); Iduno's own compiler — not an LLM — tur