AllocContext
Portfolio-aware crypto context for AI agents. Discover holdings, market, sentiment, macro, and regime; optional allocation analysis. Hosted streamable HTTP MCP with x402 on Base (USDC/EURC). Privacy: pass-through only — keys never stored. Also on PyPI: pip install "alloc-context\[mcp]". Integration: https://github.com/negillett/alloc-context/blob/main/docs/agent-integration.md
README
AllocContext
mcp-name: io.github.negillett/alloc-context
Portfolio-aware crypto context for agents — discover holdings, market, sentiment, macro, and regime; optional allocation analysis. Deterministic JSON over MCP with x402 pay-per-call on Base.
Privacy: nothing stored · one-time read-only · pass-through only — your keys and portfolio never persist on our servers. See USE.md.
Quick start (Cursor)
1. Install
pip install "alloc-context[mcp,hosted]"
# From source: pip install -e ".[mcp,hosted]"
2. User config
Copy config/user.example.yaml to
~/.config/alloc-context/user.yaml. Add read-only exchange keys for portfolio
discovery (optional) and an x402 payer for hosted market context. See
user-config.md.
3. MCP config
Add to your Cursor mcp.json (or project .cursor/mcp.json):
{
"mcpServers": {
"alloc-context": {
"command": "alloc-context",
"args": [
"mcp",
"--user-config",
"/Users/you/.config/alloc-context/user.yaml"
]
}
}
}
Use an absolute path for --user-config. Example:
cursor-mcp-bridge.example.json.
4. Ask your agent
Call get_context_bundle for a full snapshot (holdings when keys are set,
market/sentiment/macro via hosted upstream). Pure math tools
(check_allocation_band, get_rebalance_plan) work without exchange keys.
Full setup guide: cursor-mcp.md. Sample responses: examples.md.
Not financial advice.
Hosted MCP
| URL | https://mcp.alloc-context.com/mcp |
| Discovery | llms.txt, x402 manifest |
| Pricing | $0.02 cached context/math · $0.05 live ingest or portfolio |
| Payment | x402 on Base — USDC or EURC |
Agents and wallets connect directly to the hosted endpoint — see agent-integration.md. The Cursor bridge above combines local portfolio reads with this upstream for market context.
MCP tools
| Tool | Purpose |
|---|---|
get_context_bundle |
Full ContextBundle — holdings, market, sentiment, macro, delta, regime; optional allocation_analysis |
get_market_context |
Sentiment, macro, ETF, breadth, and market fields (no portfolio) |
get_context_at |
Saved snapshot from ingest history at a given as_of |
get_context_delta |
Notable shifts between two saved snapshots |
get_rebalance_plan |
USD rebalance moves from allocation, target, and NAV |
check_allocation_band |
Drift vs target and whether allocation is outside the band |
check_allocation_bands |
Batch band checks for multiple target scenarios |
get_portfolio_state |
Live NAV and holdings from Kraken or Coinbase |
Market context is holdings-scoped: band assets (BTC/ETH) use OHLC bars; alt
holdings (e.g. HYPE) use quote snapshots when cached. The bridge auto-scopes
assets from your portfolio (symbols only upstream). See
context-bundle.md#market-coverage.
See mcp.md for arguments, pricing, and resources.
Self-host and development
Run ingest and MCP entirely on your machine — no x402 upstream required.
See self-hosting.md (self_host: true in user config)
or local-dev.md for the dev stack.
git clone git@github.com:negillett/alloc-context.git
cd alloc-context
python3.11 -m venv .venv && source .venv/bin/activate
pip install -e ".[dev]"
cp .env.example .env
cp config/config.example.yaml config/config.yaml
python -m alloccontext ingest --dry-run
python -m alloccontext rollup --scope daily --stdout
pytest
| Command | Purpose |
|---|---|
python -m alloccontext ingest |
Pull configured sources → SQLite |
python -m alloccontext rollup --scope daily --stdout |
ContextBundle JSON (facts) |
python -m alloccontext status |
Per-source ingest ages, snapshots, MCP /health |
alloc-context mcp |
MCP server (stdio or HTTP) |
HTTP MCP + x402: mcp-http.md. CLI entry point:
alloc-context (same as python -m alloccontext).
Documentation
| Document | Purpose |
|---|---|
| docs/cursor-mcp.md | Cursor stdio MCP (bridge default) |
| docs/user-config.md | Bridge user.yaml reference |
| docs/mcp.md | MCP tools and x402 |
| docs/agent-integration.md | Paid HTTP MCP + Bazaar for agents |
| docs/examples.md | Sample tool JSON (redacted) |
| docs/context-bundle.md | ContextBundle schema |
| docs/USE.md | Self-host vs hosted MCP (plain language) |
| docs/mcp-http.md | HTTP MCP + x402 setup |
| docs/mcp-discovery.md | Bazaar and agent discovery |
| docs/self-hosting.md | Optional Linux/systemd ingest + MCP |
| docs/local-dev.md | Local internal MCP + dev ingest |
| docs/architecture.md | Pipeline and trust boundaries |
| docs/data-sources.md | Ingest sources |
| docs/distribution.md | GitHub, PyPI, MCP Registry, directories |
| docs/publishing.md | Release workflow and version bumps |
| docs/security-ci.md | CI coverage, Bandit, and pip-audit gates |
Contributing
GitHub Issues are welcome for bugs, schema feedback, and MCP API suggestions. Unsolicited pull requests are not expected — see CONTRIBUTING.md.
License
Elastic License 2.0 — source-available, self-host friendly. See docs/USE.md for plain-language allowed uses.
Official hosted MCP: https://mcp.alloc-context.com/mcp
推荐服务器
Baidu Map
百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。
Playwright MCP Server
一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。
Magic Component Platform (MCP)
一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。
Audiense Insights MCP Server
通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。
VeyraX
一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。
Kagi MCP Server
一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。
graphlit-mcp-server
模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
Exa MCP Server
模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。