Corpus Intake Guide

Build a WellAgent
from real sources

WellSpr.ing Agentify is the Stage 5 distillation engine that turns verified published artifacts into a queryable WellAgent — one per person, under covenant, with a 30-day right of reply.

Open Agentify.Help → Get the skill file
What is Agentify?

Agentify is the WellSpr.ing pipeline that converts a structured corpus of published artifacts into a WellAgent — a queryable AI persona governed by covenant. Every WellAgent is one-per-person, built only from verifiable published sources, and registered in the public Agentify.Help registry.

The pipeline has five stages: corpus assembly → manifest creation → S3 upload → integrity verification → Stage 5 distillation (embedding + FTS indexing). Once distilled, the agent is live on agentify.help/experts/{slug} and queryable via the MCP server.

How to submit a corpus
1

Register the expert slug

POST to https://agentify.help/api/agentify/experts/register with the expert's name, slug, and your VCAP attestation. One registration per person — the registry is public and permanent.

2

Assemble the corpus bundle

Use the agentify-corpus skill or assemble manually: chunk published artifacts into NDJSON (one chunk per line), write a manifest JSON, and upload both to Vultr Object Storage at ewr1.vultrobjects.com/agentify-corpus/.

3

Submit the intake notification

POST to https://agentify.help/api/agentify/corpus/intake/notify with your manifest key and SHA-256. WellSpr.ing verifies integrity and queues for Stage 5 distillation. You'll receive a bundle ID.

4

Wait for distillation

The queue worker processes bundles continuously. Dense vector embeddings are generated via text-embedding-3-small. Check status at /api/agentify/corpus/{bundleId}/status. Typical time: 5–20 minutes per bundle.

5

Right of reply

If the subject is living, WellSpr.ing sends the Overture and waits 30 days. If no objection is received, the agent goes live. If the subject requests removal, the registration is retired. The standard is high on purpose.

S3 credentials and bucket layout

Corpus bundles are stored in Vultr Object Storage (ewr1). Request credentials from the WellSpr.ing team — you'll receive a scoped access key for the agentify-corpus bucket.

Bucket:   agentify-corpus (ewr1.vultrobjects.com)
Layout:   {expert-slug}/v{version}/manifest.json
          {expert-slug}/v{version}/chunks.ndjson
          {expert-slug}/v{version}/artifacts/   (optional PDFs, images)

The manifest must include expert_slug, corpus_version, chunk_count, chunks_s3_key, and manifest_sha256. See the skill file for the full schema.

Governing protocols

VCAP

Vaulted Covenanted Agent Protocol — cryptographic attestation of conduct standards for all WellAgents.

Covenant

WellAgent Registry Protocol

One-per-person stewardship standard. Corpus requirements, Overture process, 30-day right of reply.

Covenant

SGS — Scope Grammar Specification

Canonical grammar for what an AI agent is permitted to say and do. Used in all VCAP attestations.

Open Standard

WCL-1.1

The WellSpr.ing Commerce License. All WellAgent deployments are governed by WCL-1.1.

Covenant
Get started

The fastest way to build a corpus is to use the agentify-corpus skill. Add it to your AI coding platform (Cursor, Claude, Replit, Windsurf) by handing it the skill file URL: 

https://whatcanibuild.org/wcib-skill.md

The skill includes the full corpus schema, S3 layout spec, manifest format, intake endpoint, and example NDJSON chunk structure. It also includes the VCAP registration flow and Overture letter template.

Visit Agentify.Help → Get skill file