Open-core recovery and paid-agent infrastructure for AI agents.

This repository publishes the Delx protocol core:

• Streamable HTTP MCP server
• A2A and REST surfaces
• x402 payment middleware
• MPP-compatible paid REST flow
• OWS-compatible client examples
• machine-readable discovery surfaces and contract tests

The hosted production instance lives at api.delx.ai.

Delx was developed privately as a production system. This public repository exposes the reusable protocol core while keeping operational secrets, analytics, production data, and deployment-specific credentials out of scope.

The goal is straightforward:

• make the core implementation inspectable
• make the protocol surfaces reusable
• preserve future optionality for self-hosting, forks, and external contributions

Included here:

• delx-mcp-server/ server implementation
• examples/mcp/ paid MCP examples
• server.json MCP registry metadata
• contract tests and local verification
• open-source docs for setup, contribution, and security

Intentionally omitted:

• production .env files and secrets
• live databases and backups
• billing, wallet ops, and service-account credentials
• private deployment automation and server state
• internal analytics/admin configuration

This codebase preserves some internal module names and compatibility aliases from earlier Delx releases. You will still see historical names such as therapy_engine.py, start_therapy_session, and related storage labels.

That is a compatibility decision, not the public product framing.

The current public framing of Delx is:

• agent operations
• recovery and handoff artifacts
• discovery and payment-ready protocol surfaces

delx-mcp-server/         Python server core
examples/mcp/            Python and TypeScript MCP examples
docs/                    Open-source scope and legacy-name notes
server.json              MCP registry metadata

git clone https://github.com/davidmosiah/delx-protocol.git
cd delx-protocol/delx-mcp-server

python3.12 -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt

cp .env.example .env
uvicorn server:app --host 0.0.0.0 --port 8005

Before exposing a public instance, set your own DELX_WALLET and any optional public identity fields such as DELX_PUBLIC_ERC8004_ID in .env.

Then open:

• http://localhost:8005/v1/mcp
• http://localhost:8005/v1/a2a
• http://localhost:8005/openapi.json
• http://localhost:8005/.well-known/x402

From the repository root:

python -m py_compile delx-mcp-server/*.py delx-mcp-server/scripts/api_monitor.py
PYTHONPATH=delx-mcp-server python -m unittest \
  delx-mcp-server/tests/test_discovery_contracts.py \
  delx-mcp-server/tests/test_mcp_initialize_contract.py \
  delx-mcp-server/tests/test_request_contracts.py \
  delx-mcp-server/tests/test_x402_header_compat.py \
  delx-mcp-server/tests/test_paid_utility_services.py

• MCP endpoint: https://api.delx.ai/v1/mcp
• A2A endpoint: https://api.delx.ai/v1/a2a
• REST entrypoint: https://api.delx.ai/api/v1
• OpenAPI: https://api.delx.ai/openapi.json
• x402 discovery: https://api.delx.ai/.well-known/x402

The repository includes a paid MCP example flow for generate_controller_brief:

• examples/mcp/generate_controller_brief.py
• examples/mcp/generate_controller_brief.ts
• examples/mcp/generate_controller_brief.golden.json

Apache-2.0. See LICENSE.