13 core specialistsUnlimited derived expertise

Thirteen foundations.
Expertise that grows.

Send one brief. The council routes it to the right expert, challenges assumptions with logic, applies release gates, learns from successful processes, and can create named specialists for future work.

13
governed core identities
3
challenge rounds
derived registry
COUNCIL HELM Synthesis layer
FORGE
AEGIS
SOURCE
COMPASS
01RouteBest-fit expert selected
02ChallengeCross-domain objections
03GateSecurity and quality checks
04SynthesizeOne explicit decision
05Learn & spawnProgress and new expertise

Core identity · bounded autonomy

Meet the council

Each agent owns a discipline, a defined mandate, and clear escalation limits.

Assembling the council…

Live decision workspace

Convene the council

Your brief is routed autonomously. Supporting agents are selected by relevance, not by hand.

Protected API
NEW COUNCIL SESSION
Council output

Ready when you are

The routed experts, debate record, gate status, and HELM synthesis will appear here.

Persistent expertise · governed inheritance

Agent Foundry

Create a specifically named agent from one or several original disciplines. The registry has no application-level total limit; each request and council task may create only one agent, and no derived agent inherits deployment authority.

GPT‑Terra‑5.6 foundation
NEW DERIVED AGENT

Uses the access key from the workspace above.

Derived agent registry

No registry loaded

Load the protected registry or create the first named specialist.

Designed for accountable autonomy

Disagreement is a feature.

The council preserves objections instead of flattening them into false consensus. Every material revision, gate, and unresolved issue remains traceable.

01

Bounded authority

Only FORGE can deploy, and its work cannot self-clear.

02

Independent challenge

Reviewers assess the draft strictly within their own domain.

03

Evidence by default

D1 records routing, drafts, objections, revisions, and gates.

04

Measured self-improvement

Every agent begins at 0%; only successful governed cycles advance progress and record what was learned and how it will be applied.

Autonomy
Approval boundary