Developers · MCP-native · agent-first

25M+ chart patterns. One MCP install. Nine canonical tools.

Drop the MCP server into Claude, ChatGPT, Cursor — any agent — and it can anchor any (symbol, date, timeframe), retrieve the cohort of 300 historical analogs, and ground every read in calibrated empirical data.

25M+indexed patterns

<300mscohort retrieval

50–0vs ungrounded agents

9canonical MCP tools

$0Sandbox tier

Get an API key →Interactive API docs MCP source →

Three lines to ship

Install once. Your agent gets nine tools and a memory layer.

Same package powers Claude Desktop, Cursor, and anything else speaking MCP. Just point it at the binary and reload.

# 1. Install the MCP server
pip install chartlibrary-mcp

# 2. Add to Claude Desktop config
{ "mcpServers": { "chartlibrary": {
    "command": "chartlibrary-mcp",
    "env": { "CHART_LIBRARY_API_KEY": "cl_your_key_here" }
} } }

# 3. Ask Claude anything anchor-shaped
> What's the read on NVDA at 1h?

Latest: chartlibrary-mcp 6.0.0·Changelog →·PyPI →

The 9 canonical tools

One MCP surface. Your agent picks the right tool for the question.

Tight surface = better tool selection. Each tool composes the others — chain a cohort_id through analyze and explain without re-running kNN.

decision_briefStart here

One-call orchestrator. Read summary flags, paraphrase in your voice.

search

Entry point. Find historical analogs of an anchor; returns a cohort_id you can chain.

cohort_analyze

The core primitive. Calibrated conditional return distribution + feature importance + regime stratification.

cohort_introspect

Slice/probe a cohort_id by macro · technical · event attributes — per-subset stats vs the full-cohort baseline.

symbol_intelligence

Layer 5 memory: per-symbol feature reliability + achieved calibration across prior analyses.

analyze

anomaly · volume_profile · crowding · correlation_shift · earnings_reaction · pattern_degradation · regime_accuracy

context

Market regime · ticker metadata · anchor metadata · system coverage. target=.

explain

Filter ranking / prose / position guidance / risk ranking from a stored cohort_id. style=.

portfolio

Multi-holding weighted distribution OR per-symbol Layer 5 memory. mode=.

report_feedback

File an error or suggestion back to the project.

REST endpoints · for non-MCP callers

Plain REST works too.

The MCP tools are wrappers over the same JSON API. If you’re building outside the MCP ecosystem, call these endpoints directly.

Method	Endpoint	Description
POST	/api/v1/decision_briefNEW	Output-first orchestrator. Returns summary block + structured cohort fields.
GET	/api/v1/agent/setups	Today's top setups pre-enriched with full-cohort stats.
GET	/api/v1/market-contextCONTEXT	One-call market awareness: SPY/QQQ/VIX + regime + sectors + crowding.
POST	/api/v1/cohort_analyze	Layer 3: cohort + feature importance + regime stratification + summary.
POST	/api/v1/cohort_compare	A/B two anchors side-by-side with delta summary.
GET	/api/v1/symbol_intelligence/{symbol}	Per-symbol Layer 5 memory — track record + feature reliability.
POST	/api/v1/search/text	Pattern similarity search. Returns cohort_id for chaining.
POST	/api/v1/analyze	Single metric drill-down (anomaly, volume_profile, crowding, etc.).
POST	/api/v1/portfolio/analyze	Batch portfolio health check (up to 20 symbols).
GET	/api/v1/narrative_pulse/{symbol}	Single-symbol FinBERT pulse + narrative-change score.
GET	/api/v1/narrative_alerts	Market-wide sentiment-price divergences.
GET	/api/v1/status	Database coverage + embedding count + date range.

OpenAPI spec →·Interactive Swagger UI →·Open in Postman →

200 free calls a day. No card.

Sandbox tier covers every solo agent on the planet. Upgrade only when you ship something that needs more.

Get an API key →See all tiers