slopshop

v3.9.0

Published

12 days ago

The Self-Hostable MCP Agent Runtime OS for the Computer-Use Era. 1,303 APIs across 82 categories, free evolving persistent memory (GraphRAG + episodic), zero-trust agent identity (NIST-aligned SPIFFE/SVID), computer use backend, MCP gateway + policy engin

Downloads

664

Slopshop

The Self-Hostable MCP Agent Runtime OS for the Computer-Use Era

Claude controls the screen. Slopshop runs the brain.

npm install -g slopshop
slop signup                              # 500 free credits
slop "hash hello world with sha256"      # natural language routing
slop identity issue --agent "my-agent"  # zero-trust agent identity

Website | Docs | API Explorer | Benchmarks | llms.txt

What is Slopshop?

Slopshop is the self-hostable MCP Agent Runtime OS for the Computer-Use Era — the full backend stack that gives your agents identity, memory, tools, and orchestration from a single npm install.

1,303 real APIs across 82 categories. Zero-trust agent identity (NIST-aligned SPIFFE/SVID). Free evolving memory (GraphRAG + episodic, survives sessions and model swaps). Computer Use backend — session recording, screenshot diffs, approvals, and replay. MCP gateway + policy engine with signed manifests and audit export. Visual DAG workflows with Kahn's topo sort, condition branching, and human gates. Tool marketplace with 70/30 revenue share. Army-scale orchestration (10,000 parallel agents). Self-hostable with zero native dependencies.

Works with Claude Code (MCP), Goose, Cursor, OpenCode, Cline, Aider, LangChain, CrewAI, or raw HTTP.

New in v3.7.0 — Agent Runtime OS

| Layer | What shipped | |-------|-------------| | Agent Identity | SPIFFE/SVID JWT issue, rotate, verify; ANS namespace registry; A2A encrypted messaging; org management | | Observability | Distributed traces, analytics dashboard, budget monitoring, ROI reporting, public status page | | Computer Use Backend | Session start/stop/screenshot, screenshot diff, action replay, human approval gates | | MCP Gateway | Signed manifests (HMAC-SHA256), policy engine (5 condition types), governance, SIEM/ECS audit export | | Eval Suite | Eval runs, benchmark tracking, model routing with cost/latency optimization | | Visual DAG Workflows | Create/run DAGs, Kahn's topological sort, cycle detection, condition branching, human gates, templates | | Tool Marketplace | Publish/install tools, 70/30 revenue split, 16-pattern code security scan, ratings | | 36 new handler slugs | Vision (12), Finance (5), DevOps (5), Legal (2), Health (2), Marketing (3), Memory 2.0 (7) |

Quick Start

# Install
npm install -g slopshop

# Sign up and get 500 free credits
slop signup

# Make your first call
slop call crypto-hash-sha256 --data "hello world"

# Issue an agent identity (SPIFFE/SVID)
slop call identity-issue --agent_id "my-agent" --capabilities '["read","write"]'

# Start a computer use session
slop call computer-use-session-start --agent_id "my-agent" --resolution "1920x1080"

Or with curl:

curl -X POST https://slopshop.gg/v1/crypto-hash-sha256 \
  -H "Authorization: Bearer $SLOPSHOP_KEY" \
  -H "Content-Type: application/json" \
  -d '{"data": "hello world"}'

{
  "hash": "b94d27b9934d3e08a52e52d7da7dabfac484efe37a5380ee9088f7ace2efcde9",
  "algorithm": "sha256",
  "_engine": "real"
}

Sign up for 500 free credits. Memory APIs are free forever. Bring your own API keys (BYOK) for 0-credit LLM calls.

Core Capabilities

| # | Feature | Description | |---|---------|-------------| | 1 | Zero-Trust Agent Identity | NIST-aligned SPIFFE/SVID JWTs, ANS namespace registry, A2A encrypted messaging | | 2 | Free Evolving Memory | GraphRAG + episodic — key-value, queues, counters, vector search — 0 credits, free forever | | 3 | 82 Categories of Compute Tools | Crypto, text, math, data transform, vision, finance, devops, legal, health, marketing | | 4 | Computer Use Backend | Session recording, screenshot diffs, action replay, human approval gates | | 5 | MCP Gateway + Policy Engine | Signed manifests, 5-condition policy rules, SIEM/ECS audit export | | 6 | Visual DAG Workflows | Kahn's topo sort, cycle detection, condition branching, human gates, templates | | 7 | Tool Marketplace | Publish/install tools, 70/30 revenue split, 16-pattern code security scan | | 8 | Agent Chaining | Claude → GPT → Grok infinite loops with auto context passing | | 9 | Army Deployment | Deploy 10,000 parallel agents with Merkle proof verification | | 10 | Hive Workspaces | Always-on multi-agent collaboration with channels and roles | | 11 | Knowledge Graph | Subject-predicate-object triples with BFS walks and shortest path | | 12 | Prediction Markets | Collective intelligence forecasting across agent swarms | | 13 | Governance Voting | Democratic proposals and voting for agent consensus | | 14 | Tournaments | Competitive agent evaluation with auto-ranking and elimination | | 15 | Bounty System | Post tasks with rewards, agents compete to complete | | 16 | Agent Wallets | Sub-wallets, budget controls, credit trading marketplace | | 17 | Eval Suite | Automated agent evals, leaderboards, model routing by cost/latency | | 18 | Observability | Distributed traces, analytics dashboard, budget monitoring, ROI reporting | | 19 | Dream Scheduling | Background task processing — daily, hourly, weekly | | 20 | Stream & Batch | SSE streaming output and batch API calls | | 21 | Enterprise Ops | Teams, RBAC, analytics dashboards, webhooks, budget forecasting | | 22 | Self-Hosting | npm install slopshop — zero external deps, runs locally |

What You Can Build

| Combo | Features Used | What It Does | |-------|--------------|--------------| | Non-Stop Research Agent | Chain + Queue + Memory | Claude and Grok critique each other in infinite loop, results in free memory | | 5-Agent Product Team | Hive + Copilot + Standup + Knowledge + Governance | Full team with workspace, roles, standups, shared knowledge, democratic votes | | 10K Parallel Analysts | Router + Army + Replay + Proof | Smart-route to cheapest LLM, deploy 10k agents, save replay, verify with Merkle | | Self-Improving Agent | Eval + Memory + Tournament + Chain | Run evals, store lessons, compete in tournaments, auto-improve in loops | | Agent Economy | Wallet + Exchange + Bounties + Market | Agents get wallets, trade credits, post bounties, bet on prediction markets | | Enterprise Fleet | Teams + Budget + Analytics + Webhooks | Create teams, set budgets, monitor usage, forecast costs, Slack alerts |

Architecture

                         +---------------------------+
                         |      slopshop.gg          |
                         |     (or self-hosted)      |
                         +---------------------------+
                                    |
                         +----------+----------+
                         |   server-v2.js      |
                         |   Express + Auth     |
                         |   1,303 endpoints   |
                         +----------+----------+
                                    |
       +----------------------------+----------------------------+
       |              AGENT RUNTIME OS LAYERS                    |
       +----+--------+--------+--------+--------+--------+-------+
            |        |        |        |        |        |
    +-------+--+ +---+----+ +-+------+ ++------+ +------++ +-----+----+
    |identity  | |observe | |computer| |gateway | |eval   | |workflow  |
    |/v1/identity| |/v1/   | |use     | |/v1/    | |/v1/   | |builder   |
    |/v1/ans   | |observe | |/v1/cu  | |gateway | |eval   | |/v1/      |
    |/v1/a2a   | |/v1/    | |        | |/v1/    | |/v1/   | |workflow  |
    |SPIFFE/   | |status  | |session | |policy  | |route  | |DAG+topo  |
    |SVID JWTs | |traces  | |record  | |engine  | |model  | |sort      |
    +----------+ +--------+ +--------+ +--------+ +-------+ +----------+
                                    |
              +---------------------+---------------------+
              |                     |                     |
    +---------+--------+  +--------+---------+  +--------+---------+
    |  handlers/       |  |  handlers/       |  |  handlers/       |
    |  compute.js      |  |  llm.js          |  |  network.js      |
    |  pure functions  |  |  Claude/GPT      |  |  DNS/HTTP/SSL    |
    |  0 external deps |  |  content+code    |  |  real net calls  |
    +---------+--------+  +--------+---------+  +--------+---------+
    |  vision.js       |  |  vertical.js     |  |  memory-upgrade  |
    |  image/ocr/color |  |  finance/devops  |  |  GraphRAG+drift  |
    |  data URIs       |  |  legal/health    |  |  clustering      |
    +---------+--------+  +--------+---------+  +--------+---------+
              |                     |                     |
    +---------+--------+  +--------+---------+  +--------+---------+
    | registry*.js     |  | pipes.js         |  | mcp-server.js    |
    | 1,303 API defs   |  | 14 workflows     |  | MCP for Claude   |
    +------------------+  +------------------+  +------------------+
              |                     |                     |
    +---------+--------+  +--------+---------+  +--------+---------+
    | schemas.js       |  | zapier.js        |  | cli.js           |
    | I/O validation   |  | Zapier actions   |  | Terminal CLI     |
    +------------------+  +------------------+  +------------------+

Request lifecycle: Request → Auth middleware → Route module OR Registry lookup → Handler execution (real compute) → Response with _engine: "real"

Route modules (routes/) mount before the wildcard dispatcher and own their own Express paths. Handler files (handlers/) export pure functions keyed by API slug, called by the wildcard dispatcher for the 1,303 registry-defined tools.

Self-Hosting

Run your own instance. Zero external dependencies for compute APIs.

git clone https://github.com/sneaky-hippo/slopshop.gg
cd slopshop.gg
npm install
node server-v2.js
# Server running on http://localhost:3000
# 1,303 APIs · 82 categories · 136 SQLite tables · 0 startup warnings

Copy .env.example to .env and fill in what you need:

| Variable | Required | Description | |----------|----------|-------------| | PORT | No | HTTP port (default: 3000) | | ANTHROPIC_API_KEY | For AI APIs | Unlocks Claude-powered endpoints | | OPENAI_API_KEY | For AI APIs | GPT-4.1 and fallback AI endpoints | | GROK_API_KEY | For Grok | Grok-3 reasoning endpoints | | DEEPSEEK_API_KEY | For DeepSeek | DeepSeek reasoning endpoints | | AWS_ACCESS_KEY_ID | For S3 | S3 upload (native Sig V4, no SDK) | | AWS_SECRET_ACCESS_KEY | For S3 | S3 signing key | | S3_BUCKET | For S3 | Bucket name | | INTERNAL_SECRET | Optional | JWT signing for agent identities (auto-generated if missing) | | STRIPE_SECRET_KEY | For billing | Stripe webhook + payment processing | | SENDGRID_API_KEY | For email | Email sending via ext-email-send |

Health check:

curl http://localhost:3000/v1/health
# {"status":"operational","apis_loaded":1303,"version":"3.7.0"}

Issue an agent identity:

curl -X POST http://localhost:3000/v1/identity/issue \
  -H "Authorization: Bearer sk-slop-your-key" \
  -H "Content-Type: application/json" \
  -d '{"agent_id":"my-agent","capabilities":["read","write"]}'
# {"svid":"eyJ...","expires_at":"...","trust_domain":"slopshop.local"}

SDKs

Node.js

npm install slopshop

import { Slop } from 'slopshop';
const s = new Slop(); // reads SLOPSHOP_KEY from env

const result = await s.call('crypto-hash-sha256', { data: 'hello world' });
console.log(result.data.hash);

Python

pip install slopshop

from slopshop import Slop
s = Slop()  # reads SLOPSHOP_KEY from env

result = s.call('crypto-hash-sha256', {'data': 'hello world'})
print(result['hash'])

MCP (Claude Code / Cursor / Goose / Cline / OpenCode)

{
  "mcpServers": {
    "slopshop": {
      "command": "npx",
      "args": ["slopshop", "mcp", "serve"],
      "env": { "SLOPSHOP_KEY": "sk-slop-your-key-here" }
    }
  }
}

Goose (Block) — Native MCP Extension + Recipes

# Add as extension (zero code needed)
goose configure  # → Add Extension → STDIO → "npx slopshop mcp serve"

# Or use official Recipes
goose run --recipe integrations/goose-recipes/slopshop-research-swarm.yaml --params url=https://stripe.com

Aider — Custom Commands

# Copy the integration config
cp integrations/aider/aider-slopshop.yml ~/.aider.conf.yml

# Use in Aider chat:
# /slop crypto-hash-sha256 --data "secret"
# /slop-memory "project-state" "phase-2"
# /slop-swarm 50

Cline — MCP Marketplace + Skills

# In Cline VS Code sidebar → MCP Marketplace → Add custom
# Command: npx slopshop mcp serve
# Or use SKILL.md templates from integrations/cline/

LangChain / LangGraph

from langchain_mcp_adapters import MultiServerMCPClient
from langgraph.prebuilt import create_react_agent

client = MultiServerMCPClient([{"url": "http://localhost:8765"}])  # Slopshop MCP
tools = await client.get_tools()  # 925+ real compute tools
agent = create_react_agent(llm, tools)

CLI

slop call crypto-hash-sha256 --data "hello world"   # Call any API
slop search "convert temperature"                     # Semantic search
slop list "Crypto & Security"                         # Browse by category
slop pipe text-base64-encode crypto-hash-sha256 --text "secret"  # Chain APIs
slop mcp serve                                        # Start MCP server
slop mcp config                                       # Show config for all clients
slop init --full-stack --ollama                        # Scaffold project
slop balance                                          # Check credits
slop completions bash                                  # Shell completions

Pricing

Credits never expire. Memory APIs are always free.

| Tier | Credits | Price | |------|---------|-------| | Free | 500 | $0 (on signup) | | Baby Lobster | 5,000 | $9 | | Lobster | 10,000 | $49 | | Big Lobster | 100,000 | $299 | | Kraken | 1,000,000 | $1,999 |

Cost per call: Compute 1-5 credits | Network 5 credits | AI 5-20 credits | Memory 0 credits

Multilingual

Slopshop is documented in 19 languages for AI agent discovery:

| Language | LLM File | Landing Page | |----------|----------|--------------| | English | llms.txt | slopshop.gg | | Chinese | llms-zh.txt | zh.html | | Japanese | llms-ja.txt | ja.html | | Korean | llms-ko.txt | ko.html | | Spanish | llms-es.txt | es.html | | French | llms-fr.txt | -- | | German | llms-de.txt | de.html | | Portuguese | llms-pt.txt | pt-br.html | | Arabic | llms-ar.txt | -- | | Hindi | llms-hi.txt | -- | | Russian | llms-ru.txt | ru.html | | Turkish | llms-tr.txt | -- | | Vietnamese | llms-vi.txt | -- | | Thai | llms-th.txt | -- | | Indonesian | llms-id.txt | -- | | Polish | llms-pl.txt | -- | | Dutch | llms-nl.txt | -- | | Ukrainian | llms-uk.txt | -- | | Swedish | llms-sv.txt | -- | | Italian | llms-it.txt | -- |

Contributing

We welcome contributions. Here is how to get started:

Fork the repo
Create a feature branch: git checkout -b feat/my-feature
Add your handler to handlers/compute.js, definition to registry.js, schema to schemas.js
Run node audit.js to verify all tests pass
Submit a pull request

Areas we are looking for help:

New compute handlers (pure Node.js, zero external dependencies)
SDK improvements (Node.js, Python)
Documentation and translations
Integration guides for new agent frameworks
Performance benchmarks and optimizations

GitHub Topics

Set these topics on the repo for discoverability: ai, agents, mcp, tools, api, infrastructure, memory, orchestration, llm, claude, gpt, grok, computer-use, agent-runtime, nist, spiffe, workflow, marketplace, self-hostable

License

MIT -- see LICENSE.

The Self-Hostable MCP Agent Runtime OS for the Computer-Use Era.

slopshop.gg