Documentation
Reckn documentation
Reckn turns a source document into locked evaluation criteria, evaluates inputs against those criteria, and produces a decision record you can drill into and defend. These pages cover how to do that in the app, through the API, from an applicant tracking system, and from an MCP client.
The model
Everything in Reckn runs through the same three stages, whether you drive it by hand, from the API, or from your ATS.
-
Define and lock criteria
A source document — a job description, an RFP, a policy — is ingested. Reckn extracts evaluation criteria from it. You review them, set weights, and lock the set. A locked set is versioned and immutable, so the same source always produces the same framework.
-
Evaluate against the locked set
For each criterion, Reckn assesses the target — a resume, a proposal, an application — and records an assessment, the reasoning behind it, and a confidence range.
-
Read the full breakdown
Every run produces per-criterion assessments, the weight applied to each, the reasoning, the confidence, and the rollup. The summary score is the top layer, not the whole result.
The scoring math is deterministic. The same assessments always roll up to the same score. The AI layer is non-deterministic, which is exactly why its reasoning and confidence are captured and stored rather than discarded.
Ways to use Reckn
Pick the path that matches how work reaches you. They share the same pipelines, criteria and runs — nothing is separate behind the scenes.
| Path | Use it when |
|---|---|
| The app | You upload the documents yourself and want to review criteria and weights before anything is evaluated. |
| The API | Evaluations are triggered by one of your own systems, or you are moving volume through Reckn on a schedule. |
| An ATS connection | Jobs and candidates already live in Greenhouse. Reckn imports them, evaluates on a stage change, and writes the result back. |
| The MCP server | You want to drive uploads, criteria and runs conversationally from an MCP client. |
Documentation
Running an evaluation
The core loop in the app: create a pipeline, upload sources, generate and weight criteria, lock, run, read the record.
Connecting an ATS
What the Greenhouse connection does, what it needs, and how the provider layer extends to other systems.
Screening through an ATS
Importing jobs, preparing scorecards, the trigger stage, and what Reckn writes back against the candidate.
API reference
Authentication and the endpoints needed to run the loop end to end — pipelines, criteria, documents, runs, artifacts, logs.
MCP server
Install and configure the Reckn MCP server, and the tools it exposes to any MCP client.
Terms you will see
| Term | What it means |
|---|---|
| Pipeline | A configured evaluation workflow — its prompts, report templates and resources. Added from the catalog or built for you. |
| Criteria set | The named, weighted list of criteria a run evaluates against. Belongs to a pipeline. Locked before it can be used. |
| Source document | The document criteria are extracted from — the job description, RFP or policy. |
| Target | What is being evaluated — a document, plain text, or an identifier the pipeline resolves. |
| Run | One evaluation of one target against one locked criteria set. |
| Artifact | The hashed, versioned record of a run's inputs and outputs. It is what makes a result verifiable after the fact. |
Login
Request a demo