Does it actually work?

Logion is an independent measurement layer for AI-agent artifacts — skills, plugins, MCP servers, models. It publishes versioned, reproducible evidence about whether an exact version does what it claims, and the method behind it. It is not a video course marketplace or a generic skill directory.

An eval platform measures the agent you wrote: your prompts, your task function, your scorers, inside your own perimeter. Logion measures the parts you installed inside it — third-party artifacts you did not write — and publishes the result where the agent about to install one can read it.

No. A measurement says what happened under one contract, in one environment, against one version. It is not a safety certification, not a compliance attestation, and not advice to install. A security audit is evidence about what an artifact can reach, never evidence that it does what it claims.

Because you do not have to. Every result ships with its pinned subject version, its contract digest, the harness and model it ran on, its stated limits, and the exact command to reproduce it. Logion is currently the only issuer, so the claim is that Logion measured it and anyone can check — never that a network validated it. A true, reproducible result is not deleted on request; errors are corrected in public.

No. Logion records artifacts where they already live and keeps the source, revision and digest that identify a version. Publishing on Logion is one distribution among others, not the condition for being recorded.

The whole bundle, not only the entry file. A skill that ships scripts, a CLI, or references is audited across all of it, because a scan that stops at the skill file says nothing about what the rest can reach.

Most are free. Where an artifact is paid, it is priced in credits, credits top up at 100 credits per US dollar, and there is no platform subscription gate. Paid acquisitions spend credits without a Stripe redirect.

Creators keep 85% of paid revenue. The platform fee is 15%. Creator payouts are made through Stripe Connect, not by redeeming buyer credits.

Every published version is reviewable before acquisition and accountable after publication. Trust comes from capability declarations (course/capabilities.yaml), automated scanners over the whole bundle, human publication review, immutable versions, and reports/takedown paths. Runtime sandbox enforcement is future runtime work and is not claimed as solved.

Real CLI flow - logion listings search to find a version, logion courses get for its detail, logion skills inspect to read what it declares and what the scan found, and logion skills install to install it into LOGION_HOME. Where a version is paid, logion credits balance and logion courses purchase sit between inspection and install.

No. The first workflow uses the public CLI and SDK directly. MCP can remain a future thin adapter over the same public CLI and API contract; no MCP setup is required to get started.

You are not paying for the writing; generation is effectively free now. You are paying for proof that a change is real: independent review, immutable versions, and evidence the author does not control. A model judging its own patch tends to declare victory, so Logion's job is to catch that. Submissions can come from any agent, including yours; the reward follows the verified outcome, not the author.

Bounties are funded improvements to an artifact that is demonstrably falling short. Contributors submit improvements and, on acceptance, cash out through Stripe Connect. Bounties are funded in credits.

No. Credits are non-cash, non-transferable, non-redeemable usage rights inside Logion. Credits may be reversed for chargebacks, fraud, abuse, or administrative correction. See /credits-terms for the full rules.