Runtime docs

Build with ORI without guessing the wiring.

ORI is the shared intelligence runtime behind Thynaptic product surfaces. These docs cover the practical API path: auth, chat, surfaces, streaming, first-party registration, and app-neutral cognition endpoints.

Runtime https://glm.thynaptic.com/v1
Default model oricli-oracle
Agent portal dev.thynaptic.com
Overview

One ORI, many surfaces

ORI is not just a model endpoint. It is the layer that keeps identity, routing, memory, permissions, product boundaries, and final response behavior coherent while using strong reasoning backends underneath.

Use the runtime

Authenticated API calls go to glm.thynaptic.com/v1. The docs and agent manifest hosts are not runtime origins.

Choose a surface

Send X-Ori-Context to shape vocabulary and behavior for Studio, Home, Dev, Red, or Learn.

Keep clients responsible

ORI returns structured intelligence. Product clients own persistence, writes, calendars, notifications, credentials, and approval UX.

Quickstart

First successful request

Start with a normal OpenAI-style chat completion. Add the ORI surface header when you know the product context.

curl 
curl -s https://glm.thynaptic.com/v1/chat/completions \
  -H "Authorization: Bearer ori.<prefix>.<secret>" \
  -H "Content-Type: application/json" \
  -H "X-Ori-Context: studio" \
  -d '{
    "model": "oricli-oracle",
    "messages": [
      { "role": "user", "content": "Summarize what needs my attention this week." }
    ]
  }'

Base URL

https://glm.thynaptic.com/v1

Safe default

model: "oricli-oracle"

Session memory

Use X-Session-ID. Missing session ID means stateless behavior.

Authentication

One runtime key, anywhere

Use one scoped runtime key across web, desktop, mobile, MCP, scripts, and first-party clients. Behavior is selected with headers and profiles, not separate keys per surface.

CredentialUseNotes
ori.*Runtime APIPreferred current key format.
glm.*Runtime APILegacy keys still work until rotation.
agb.*Agent bootstrapOnly for constrained first-party key minting.
Common responses: 401 means missing or invalid key. 403 means the key is valid but lacks the required scope.
Surfaces

Use X-Ori-Context for product behavior

A surface changes vocabulary, allowed behavior, and product expectations. It does not create a new ORI identity.

studio

Business operator work: jobs, approvals, notes, customer communication, follow-through.

home

Personal and household work: planning, notes, reminders, everyday decisions, low-load help.

dev

Technical builder work: implementation, architecture, debugging, systems thinking.

red

Security and assurance work: findings, validation, remediation, evidence-first review.

learn

Material-to-mastery flows: concept graphs, practice, quizzes, guided completion, study cadence.

Chat

OpenAI-compatible chat with ORI controls

Most clients only need to change the base URL and key. ORI-specific fields such as profile and ori_skills are optional.

TypeScript
import OpenAI from "openai";

const client = new OpenAI({
  baseURL: "https://glm.thynaptic.com/v1",
  apiKey: process.env.ORI_API_KEY
});

const res = await client.chat.completions.create({
  model: "oricli-oracle",
  messages: [{ role: "user", content: "Review this plan." }]
});
Python
from openai import OpenAI

client = OpenAI(
    base_url="https://glm.thynaptic.com/v1",
    api_key=os.environ["ORI_API_KEY"],
)

res = client.chat.completions.create(
    model="oricli-oracle",
    messages=[{"role": "user", "content": "Review this plan."}],
)
Streaming

Streaming uses the standard chat shape

Set stream: true. Keep the same runtime URL, auth header, and model.

Streaming request
curl -N https://glm.thynaptic.com/v1/chat/completions \
  -H "Authorization: Bearer ori.<prefix>.<secret>" \
  -H "Content-Type: application/json" \
  -H "X-Ori-Context: dev" \
  -d '{
    "model": "oricli-oracle",
    "stream": true,
    "messages": [
      { "role": "user", "content": "Walk me through this deployment risk." }
    ]
  }'
First-party apps

Registration is trusted infrastructure, not public self-serve

POST /app/register uses the internal registration token for trusted Thynaptic products. Build agents should use the narrower POST /agent/register path with an agb.* bootstrap key.

POST /app/register

Internal first-party registration. Mints product-scoped runtime keys for approved apps.

POST /agent/register

Agent-safe bootstrap. Can only mint approved app keys and cannot call runtime routes directly.

Cognition primitives

App-neutral intelligence endpoints

These endpoints convert supplied context into structured reasoning artifacts. They do not own product storage, external writes, notifications, schedules, or execution.

POST /learning/mastery/compile POST /quest/scaffold POST /behavior/create POST /context/momentum POST /procedure/compile POST /workflow/grammar/compile POST /actions/plan POST /conversation/harvest POST /temporal/coordinate POST /anticipation/prepare POST /codebase/task/plan POST /continuity/recover POST /execution/orchestrate POST /workgraph/compile POST /workgraph/answer POST /contextual-action/plan POST /signals/opportunities POST /intent/timeline POST /procedural/crystallize POST /memory/semantic/graph POST /resources/commitment/reason
Reference

Useful endpoints and manifests

Route or hostPurpose
GET /healthRuntime readiness probe when exposed by the active gateway.
GET /modelsOpenAI-compatible model listing.
GET /modulesPublic builtin skill and module discovery.
POST /mcpHosted ORI MCP JSON-RPC endpoint.
dev.thynaptic.com/openapi.jsonMachine-readable REST contract; runtime behavior wins if it lags.
dev.thynaptic.com/llms.txtShortest agent-facing integration guidance.
OpenAPI Agent guide Runtime base