Morpheus Hackathon Kit

An open-source (MIT) kit for building a hackathon product that is worth building, on Morpheus decentralized inference. Morpheus is an OpenAI-compatible API in front of a peer-to-peer network of model providers. One key and one base URL give you 250+ models, including frontier, open-weight, :web search and embedding models.

Agents: start at llms.txt. Humans: start at MORPHEUS.md for every official link.

Why build on Morpheus

What you get	Why it matters
One key, one OpenAI-compatible base URL, 250+ models from many model families	Swap models with a one-line change. You are not tied to one AI company's API
Independent providers compete on an open marketplace	Inference is supplied by many operators, not one host. Anyone can become a provider
Pay with card, USDC, or staked MOR	Inference can be funded by stake instead of a card (stakemor.com)
A TEE provider path with attestation	Build where confidential inference matters; check which providers attest before you claim it (TEE overview)
Open-source node software	Run your own consumer node and call the network directly (morpheus/NODE.md)
Builder subnets	Keep building after the event and attract MOR stake toward your project

Who it's for

You are	Start here
Building an app or product	The Inference API: quick start below
Building an agent	Morpheus Skill, or point your agent at your own node (agents)
Running models or spare API capacity	Become a provider
Holding or staking MOR	Builders and capital

Start with the problem, not the code

A product is incomplete until you have mapped product-market fit, the pain, and the cost of not solving it. Approach the build as delivering value, economic or otherwise, in a form that can last. Before you write product code, answer the four product risks from Marty Cagan's product-discovery approach:

Risk	Question
Value	Will the buyer choose this?
Usability	Can they use it without help?
Feasibility	Can we build it in time, on Morpheus?
Viability	Does it hold up as a business or a sustained effort?

Read DISCOVERY.md, then run the skills in order: /market-scan → /pain-map → /gtm-wedge → /opportunity (GO / PIVOT / STOP) → /scaffold → /poc.

Quick start (5 minutes)

1. Create an account and an API key at app.mor.org. Keys start with sk-.
2. Copy .env.example to .env, paste your key, and load it into your shell. Never commit .env.

   cp .env.example .env        # then edit .env
   set -a; source .env; set +a

3. Check your key and pick models:

   ./morpheus/preflight.sh "glm-5.3-flash" "Claude Sonnet 5.5"

   The script prints latency, a reply check and tool-calling support for each model.
4. Run an example (one chat call and one tool call, with retry and fallback):

   pip install -r examples/python/requirements.txt && python examples/python/main.py
   npm install --prefix examples/node && npm start --prefix examples/node

   Or open the repo in GitHub Codespaces, with MORPHEUS_API_KEY set as a Codespaces secret. Everything installs automatically.
5. Point any OpenAI SDK at https://api.mor.org/api/v1.
6. Optional: run claude in this folder. Claude Code loads the skills in .claude/skills/ automatically. Fill in the Brief in CLAUDE.md first.

Python

from openai import OpenAI
import os

client = OpenAI(base_url="https://api.mor.org/api/v1", api_key=os.environ["MORPHEUS_API_KEY"])
r = client.chat.completions.create(
    model=os.getenv("MODEL_PRIMARY", "glm-5.3-flash"),
    messages=[{"role": "user", "content": "Say hello from Morpheus."}],
)
print(r.choices[0].message.content)

TypeScript

import OpenAI from "openai";

const client = new OpenAI({ baseURL: "https://api.mor.org/api/v1", apiKey: process.env.MORPHEUS_API_KEY });
const r = await client.chat.completions.create({
  model: process.env.MODEL_PRIMARY ?? "glm-5.3-flash",
  messages: [{ role: "user", content: "Say hello from Morpheus." }],
});
console.log(r.choices[0].message.content);

curl

curl -s https://api.mor.org/api/v1/chat/completions \
  -H "Authorization: Bearer $MORPHEUS_API_KEY" -H "Content-Type: application/json" \
  -d '{"model":"glm-5.3-flash","messages":[{"role":"user","content":"Say hello from Morpheus."}]}'

List the live models with curl -s https://api.mor.org/api/v1/models. Model IDs are case-sensitive and can include spaces.

Choose a lane

Lane	Best for	Start here
Gateway API	Most teams. Zero infrastructure	Quick start above, then morpheus/SETUP.md
Template	A working app in minutes	Template table in MORPHEUS.md
Your own node	Teams who want to call the network directly, without the gateway	morpheus/NODE.md

What is in this repo

File	Purpose
DISCOVERY.md	The four risks, the pain / cost-of-inaction / fit maps, and further reading
CLAUDE.md	Agent rules plus a brief to fill in: event, theme, judging, approver
AGENTS.md	Practice for people and agents: research, repo, marketing, outreach, safety
.claude/skills/	16 skills for Claude Code (or any agent that reads skill files): discovery → build → ship → pitch
graphs/	Execution graphs (DAGs) that chain the skills with gates and approvals. Run with /run-graph
examples/	Runnable Python and Node examples on Morpheus
.devcontainer/	Open the repo in GitHub Codespaces with everything installed
llms.txt	Machine-readable index of every Morpheus resource, for agents to sweep
MORPHEUS.md	Human index: docs, API, templates, agents, nodes, providers, known gaps, community
morpheus/SETUP.md	Gateway setup: what the API has, what it lacks, how to pick models, how to route coding agents
morpheus/NODE.md	Run your own consumer node, safely
morpheus/preflight.sh	Tests a key, models, latency and tool calls
morpheus/no-other-vendors.sh	Pre-commit / CI check that your code calls no other AI vendor directly
.env.example	Environment variables the scripts and snippets use

Good practice

1. Put all model calls behind one llm() function. Read the model IDs from .env, so a model swap is a one-line change.
2. Pick a fallback model, also on Morpheus, and test it before demo day.
3. Log the model, the endpoint and the latency of each call, but not the prompts. On stage, that log proves where your inference runs.
4. Thinking models can take more than a minute before the first token. For a live demo, use a fast model.
5. No secrets in git, chat or prompts. Run morpheus/no-other-vendors.sh as a pre-commit hook:

   ln -s ../../morpheus/no-other-vendors.sh .git/hooks/pre-commit

After the hackathon

Keep building: register a builder subnet to attract MOR stake to your project, or fund your inference with staked MOR at stakemor.com. The known gaps list is a good place to find a project worth continuing.

Help and contributing

Ask questions in this repo's Discussions. Report kit bugs as issues. Pull requests are welcome. For the API itself, see the support links in MORPHEUS.md.

License

MIT.