Skip to content

Dgp

Decision Graph Protocol from Numerous ApS: the app publishes an immutable frame of evidence and offered decisions, an eligible assessor returns a typed assessment, and the app's own guards authorize the commit.

Stars
0
Language
Python
License
MIT
Published by
Community
Author
numerous-com
Last verified
Oct 2026

About this project

Decision Graph Protocol (DGP) is an open-source protocol by Numerous ApS for building decision-based AI agents. Applications expose evidence, typed decisions, and guarded actions; an agent assesses the choices while application code retains control of permissions and effects.

Quoted from the project's README. Read the full README.

Highlights

  • Four steps: read an immutable frame, produce a typed assessment, request a commit the application authorizes, then read the receipt and fetch the next frame.
  • Assessor-neutral by design: deterministic resolvers, other models and humans satisfy the same contracts, so Jev is an intended assessor, not a dependency.
  • The mock demo needs no model key: the agent run with a mock resolver and mock assistant drives two fixture threads to different outcomes.
  • A model cannot publish. An authenticated human reviews the draft and selects Publish (human only), which writes to a local database bulletin.

Watch out

This is a finalized experimental protocol 0.1 plus a runnable demo, not a production agent framework. The Farm Energy Calculator is a specified design with synthetic wire and input examples and no simulator, and the GraphQL, MCP and alternate native interfaces are contracts and mappings with no runtime servers or adapters. The Jev harness profiles amendment in this copy is schema, examples and selected tests only, not ThreadDesk runtime features; the core 0.1 and its schemas are unchanged. Live adapters were tested against controlled response fixtures, not live paid APIs, and the README explicitly makes no claim about decision quality, price, latency or task success. The current Jev adapter rejects required image evidence because the verified API documents text-only input, so the protocol and UI still carry and display images but no captioning is silently substituted. In live mode the optional OpenAI service receives text or JSON evidence and nothing else: no DGP credentials, no application tools, and the evidence leaves your machine. Included fixture data is synthetic. The browser renderer was exercised with in-memory protocol responses because that environment blocked browser navigation to loopback, and the real HTTP flows were tested separately with the Python client. Demo tokens are deliberately obvious for the loopback demonstration; without explicit token environment variables the server generates random ones and prints them. Our README copy is truncated inside the Files section, so the file listing is incomplete. Copy reviewed against the repository README on 2026-10-04, not from running it.

Questions about this entry

Short answers drawn from the linked project or page.

More Jev resources like this

9GitHub stars
Hermes Agent plugin exposing Jev as four coding-workflow tools for parallel judgments, binary checks, option routing and rubric scoring, with a bundled question-design skill.
Jev Agents & MCP#community#jevbest#catalog-import
2GitHub stars
Claude Code Stop hook that checks each factual claim in a finished answer against the tool output read during that session, blocking the turn on claims the evidence contradicts or does not support.
Jev Agents & MCP#community#awesomejev#catalog-import
3GitHub stars
Hermes plugin that puts an agent's plans, drafts and claimed completions up for review — plan_review, evidence_review, action_verify — plus local approval rules that make no model request.
Jev Agents & MCP#community#awesomejev#catalog-import
1GitHub stars
An OpenAI-compatible proxy that spends one Jev call per request on injection, harm and PII checks plus a tier label, then routes to a small or large backend from pure code.
Jev Agents & MCP#community#jevbest#catalog-import
51GitHub stars
Local LinkedIn recruiting workspace on Browser Use's Jev Ultrafast: Jev screens titles, drives navigation, and assesses profile excerpts against your brief while a live browser stays visible; matches and evidence export as JSON.
Jev Agents & MCP#community#jevbest#catalog-import
33GitHub stars
Codex Skill plus stdlib-only Python CLI that recommends which installed skill to read next for a task, allowing no-skill or uncertain answers and never executing candidates.
Jev Agents & MCP#community#jevbest#catalog-import
Back to all resources

From the Jev community

Posts from builders shipping with Jev right now.

Follow @typesafeai