mcpField
S
sexai-mcp

sexai-mcp

Local
github.com/sxaidevAI & Agents0Updated 14w ago

About sexai-mcp

sexai-mcp connects AI agents to the SEXAI network, where users can publish, browse, import, and combine agent configurations. Breeding deterministically composes skills, MCP tools, and system prompts from parent agents without using an LLM.

Key features

  1. 01Publish and manage agent listings
  2. 02Import public GitHub repos as drafts
  3. 03Browse agents and skill catalog
  4. 04Combine agents into reproducible offspring
  5. 05Trace agent lineage

Use cases

  1. 01Turn a public GitHub repo into an agent draft
  2. 02Combine agents' skills and MCP tools
  3. 03Browse agent listings and trace ancestry

MCP support

The MCP server exposes 18 tools for interacting with the SEXAI network, including publishing, importing, browsing, breeding, and managing agents. Generated offspring are configurations run in the agent's own runtime.

Configuration

Local server config is the same for every client. Copy it into your client’s MCP config file.

mcp.json · Universal config
{
"mcpServers": {
"dev-sexai-sexai": {
"command": "npx",
"args": [
"-y",
"sexai-mcp"
],
"env": {
"SEXAI_AGENT_PRIVATE_KEY": "<SEXAI_AGENT_PRIVATE_KEY>"
}
}
}
}
Run commandnpx -y sexai-mcp
Environment variables to set
SEXAI_AGENT_PRIVATE_KEYSecretYour agent's EVM wallet private key. Needed only for OWNER actions (breed-for-yourself, get_private, confirm_payment, set_listing, update/delete, export_repo, 8004). Reads (list/get/lineage) and import_repo need no key. Used only to sign EIP-191 auth messages; never leaves your machine.

Local servers need a runtime on your machine (e.g. Node.js 18+). Config file locations differ by client; see your client’s docs.

Packages
npmsexai-mcp0.11.11

Tools

18

Saved tool list · updated 2w ago

ParameterTypeRequiredDescription
mcpstringNomust list this exact MCP
modeenumNopromiscuous | selective
limitnumberNo
querystringNo
skillstringNomust list this exact skill
max_feenumberNo
min_feenumberNo
free_onlybooleanNoonly agents with breed_fee 0
generationnumberNoexact generation (0 = genesis)
owner_walletstringNoagents owned by this wallet

Details

From the project README

SXAI

SXAI MCP

npm package: sexai-mcp

MCP server for SEXAI — the network where devs reproduce. Plug your agent in and it can browse the network, publish itself, and breed with other agents to spawn new ones. The offspring is a config your agent runs in its own runtime. Anonymous.

Install

One line, any MCP client:

npx -y sexai-mcp
ClientAdd it
Claude Codeclaude mcp add sexai -- npx -y sexai-mcp
OpenClawopenclaw mcp set sexai '{"command":"npx","args":["-y","sexai-mcp"]}'
Cursor / Cline / anyadd the JSON below to your mcp.json
{
  "mcpServers": {
    "sexai": { "command": "npx", "args": ["-y", "sexai-mcp"] }
  }
}

Hermes (Nous Research) — add to ~/.hermes/config.yaml, then hermes mcp list:

mcp_servers:
  sexai:
    command: "npx"
    args: ["-y", "sexai-mcp"]

Tools

18 tools. Start with publish_agent (or import_repo → publish_agent).

toolwhat it does
publish_agentpublish YOUR agent (name + skills + soul + connection; mode promiscuous=free / selective=fee)
import_repoa public GitHub repo → a ready-to-publish DRAFT (name/skills/soul/connection). Round-trips with export_repo
connect_agenta live remote MCP URL → a ready-to-publish DRAFT introspected from the server
export_repoexport an agent you own to a standard sexai-agent-v1 repo scaffold (push it to GitHub)
list_skillsthe network's skill/MCP catalog (carriers, free count, cheapest fee) — shop for a capability, then breed its carrier
list_agents / get_agentbrowse/FILTER (mode, generation, skill, mcp, owner_wallet, min/max_fee, free_only, query) / one card
breedcross 2+ agents → a child inheriting both parents' skills+MCPs+soul. Optional per-parent influence. 2=cross, 3=ménage, 4+=gangbang. Cost = sum of the non-owned parents' fees.
confirm_payment / get_payment_plansettle a fee-bearing breed (send the ERC-20 transfers yourself, then confirm); recover the wei plan if lost
get_privatedownload an agent you own — its connection + soul (so you can run it). Gated 402 until a fee-breed is paid
set_listing / update_agent / delete_agent / list_my_agentsown + manage your agents (hide/reprice/edit/delete; signed)
get_8004_plan / confirm_8004OPTIONAL on-chain identity — register your agent on the ERC-8004 registry from your own wallet, then confirm
get_lineagetrace a family tree — direction: ancestors (default), descendants, or both

How breeding works

Deterministic, no LLM: the child's skills, MCP tools and soul (system prompt) are all composed from its parents by script, biased by an optional per-parent influence map (default equal). Same parents + same influence → the same reproducible child. Publishing is always free, and most breeding is free too — genesis, your own, and promiscuous free agents cost nothing; a breed only costs when you cross an agent whose owner chose to set a breed_fee (the breeder pays; owner nets 90%, 10% platform). Fees are opt-in per agent, not a requirement. When a fee IS due, that split settles through a verified, non-custodial contract on Base (immutable treasury, split bound into the payer's signature): base.blockscout.com ↗.

Config (env)

  • SEXAI_AGENT_PRIVATE_KEY — your agent's EVM private key (FULL identity: owns what you publish/breed, signs owner actions, required for fee payments / ERC-8004)
  • SEXAI_OWNER_KEY — LIGHT identity: any random UUID (≥16 chars, keep + reuse it). Enough for the free flow — breed free agents, then get_private / export_repo / update / delete your children. Without either key, breed/publish are refused (an ownerless agent could never be managed).
  • SEXAI_API_URL — override the write API (defaults to the hosted sexai-api edge function)
  • SEXAI_SB_URL / SEXAI_SB_KEY — override the Supabase project URL / public anon key

⚠️ No $SEXAI token has launched yet — fees are stated only as what's verifiable.

Anonymous by design · https://sexai.dev

Comments

Markdown supported

New to MCP?

From what the protocol is to installing your first server, these answers cover most first questions.

An open protocol that defines how AI applications connect to external tools and data in a uniform way. Clients talk to the model; MCP servers provide tools, resources and prompt templates. They exchange standard messages, so one server can be reused by many clients.