About blockchainacademics
blockchainacademics is a crypto MCP server for AI agents, covering editorial content, market and on-chain data, sentiment, and research tools. Responses include citations, timestamps, and source hashes for provenance.
Key features
- 01Search news and editorial content
- 02Query prices, markets, and on-chain data
- 03Analyze sentiment and proprietary indicators
- 04Generate research reports and models
- 05Check security, regulation, and fundamentals
- 06Return source attribution with responses
Use cases
- 01Research a token or protocol
- 02Review market and on-chain changes
- 03Assess sentiment, risk, and security
- 04Generate due diligence in agent workflows
MCP support
It exposes crypto data, content, and research tools to AI agents through MCP. Responses include citation URLs, timestamps, and source hashes.
Configuration
Local server config is the same for every client. Copy it into your client’s MCP config file.
npx -y @blockchainacademics/mcpBCA_API_KEYSecretOptional. When unset, the server uses a public demo key (10 marquee tools, 100/day shared cap). Set for the full 104-tool surface — free tier 2,000/mo at https://brain.blockchainacademics.com/signup.BCA_API_BASEOverride the API base URL. Defaults to https://api.blockchainacademics.com.Local servers need a runtime on your machine (e.g. Node.js 18+). Config file locations differ by client; see your client’s docs.
Tools
99Saved tool list · updated 1w ago
| Parameter | Type | Required | Description |
|---|---|---|---|
| limit | number | No | Max results (default 10, max 50). |
| query | string | Yes | Full-text search query (1-512 chars). |
| since | string | No | ISO 8601 date; return articles published on or after this timestamp. |
| topic | string | No | Topic filter (e.g. 'regulation', 'defi'). |
| entity | string | No | Entity slug filter (e.g. 'ethereum', 'circle'). |
Details
From the project README@blockchainacademics/mcp
The canonical crypto MCP server for AI agents. 105 tools spanning 3,501+ editorial articles, 200+ entity dossiers, 43 academy lessons, aggregated market + on-chain data, proprietary indicators, and async agent-backed generation — all accessible as MCP tools your AI agent can call natively.
Tool categories (v0.4.0)
| Category | Count | Examples |
|---|---|---|
| Content & corpus | 7 | search_news, get_article, get_entity, get_explainer, list_entity_mentions, get_as_of_snapshot |
| Market data | 4 | get_price, get_ohlc, get_market_overview, get_pair_data |
| On-chain | 4 | get_wallet_profile, get_tx, get_token_holders, get_defi_protocol |
| Sentiment | 3 | get_sentiment, get_social_pulse, get_fear_greed |
| Proprietary indicators | 6 | get_coverage_index, get_narrative_strength, get_sentiment_velocity, get_editorial_premium, get_kol_influence, get_risk_score |
| Agent-backed (async, Pro+) | 6 | generate_due_diligence, generate_tokenomics_model, summarize_whitepaper, translate_contract, monitor_keyword, get_agent_job |
| Directories | 13 | list_stablecoins, list_yields, list_aggregators, list_mcps, list_vcs, list_jobs, build_custom_indicator, … |
| Fundamentals | 6 | get_tokenomics, get_audit_reports, get_team_info, compare_protocols, check_rugpull_risk |
| Chain-specific | 4 | get_solana_ecosystem, get_l2_comparison, get_bitcoin_l2_status, get_ton_ecosystem |
| Markets microstructure | 5 | get_funding_rates, get_options_flow, get_liquidation_heatmap, get_exchange_flows, predict_listing |
| Narrative / meta | 5 | track_narrative, get_ai_agent_tokens, get_depin_projects, get_rwa_tokens, get_prediction_markets |
| Regulatory | 4 | get_regulatory_status, track_sec_filings, get_mica_status, get_tax_rules |
| Security | 4 | check_exploit_history, check_phishing_domain, get_bug_bounty_programs, scan_contract |
| Memes / degen | 4 | track_pumpfun, track_bonkfun, check_memecoin_risk, get_degen_leaderboard |
| Services (revenue) | 3 | book_kol_campaign, request_custom_research, submit_listing |
| History time-series | 4 | get_history_prices, get_history_sentiment, get_history_correlation, get_history_coverage |
| Compute / AI crypto | 2 | get_compute_pricing, get_ai_crypto_metrics |
| Corpus meta | 7 | list_entities, get_topic, search_academy, get_trending, get_unified_feed, list_sources, get_recent_stories |
| Memos + theses + social | 6 | list_memos, get_memo, list_theses, get_thesis, get_social_signals, get_social_signals_detail |
| Currencies | 2 | list_currencies, get_currency_feed |
Every tool response carries cite_url (with UTM src attribution), as_of (time-travel snapshot), and source_hash (content integrity).
Tools for unreleased integrations (Twitter social pulse, DEXTools, bonk.fun, Bittensor) return {status: "integration_pending", reason, eta} rather than failing. The MCP server is fully usable today.
Why
LLMs hallucinate about crypto. BCA ships ground-truth editorial content with full attribution. Plug this MCP server into Claude Desktop, LangChain, Eliza, or any MCP-compatible agent and your model queries the BCA corpus like any other tool — with citations, timestamps, and source hashes on every response.
Install
Claude Desktop (zero-config demo)
The package ships with a public demo key baked in. No signup needed to try it. Add to claude_desktop_config.json (macOS: ~/Library/Application Support/Claude/claude_desktop_config.json):
{
"mcpServers": {
"blockchainacademics": {
"command": "npx",
"args": ["-y", "@blockchainacademics/mcp"]
}
}
}
Restart Claude Desktop. The demo path unlocks 10 marquee tools (get_price, get_trending, get_fear_greed, get_market_overview, search_news, get_sentiment, get_entity, get_explainer, get_recent_stories, get_topic) with real data and real citations. Shared rate cap: 100 calls/day globally + 20/day per IP.
Claude Desktop (full 105 tools)
Get a free key at brain.blockchainacademics.com/signup (2,000 calls/month per user) and add the env block:
{
"mcpServers": {
"blockchainacademics": {
"command": "npx",
"args": ["-y", "@blockchainacademics/mcp"],
"env": { "BCA_API_KEY": "bca_live_xxxxxxxxxxxxxxxx" }
}
}
}
Restart Claude Desktop. All 105 tools appear in the tool picker.
Programmatic (LangChain, Eliza, custom agents)
npm install @blockchainacademics/mcp
import { Client } from "@modelcontextprotocol/sdk/client/index.js";
import { StdioClientTransport } from "@modelcontextprotocol/sdk/client/stdio.js";
const transport = new StdioClientTransport({
command: "npx",
args: ["-y", "@blockchainacademics/mcp"],
env: { BCA_API_KEY: process.env.BCA_API_KEY! },
});
const client = new Client({ name: "my-agent", version: "0.0.1" }, { capabilities: {} });
await client.connect(transport);
const res = await client.callTool({
name: "search_news",
arguments: { query: "stablecoin regulation", limit: 5 },
});
console.log(res.content);
See examples/research-agent.ts for a full tool-chaining loop.
Tools
search_news
Full-text search across 3,501+ editorial crypto articles.
| arg | type | required | description |
|---|---|---|---|
query | string (1-512) | yes | Search query |
entity | string | no | Entity slug filter (e.g. ethereum) |
since | ISO 8601 | no | Earliest publish date |
topic | string | no | Topic filter (e.g. regulation) |
limit | number (1-50) | no | Default 10 |
Example: { "query": "circle IPO", "since": "2026-01-01T00:00:00Z", "limit": 5 }
get_entity
Fetch a canonical entity dossier. Provide exactly one of:
| arg | type | description |
|---|---|---|
slug | string | e.g. "vitalik-buterin", "circle" |
ticker | string | e.g. "ETH", "SOL" (case-insensitive) |
Aliases resolve automatically (CZ → changpeng-zhao, Maker → makerdao, BSC → bnb-chain, …).
get_explainer
Fetch a canonical academy lesson. Provide exactly one of:
| arg | type | description |
|---|---|---|
slug | string | Lesson slug, e.g. "what-is-a-blockchain" |
topic | string | Topic keyword that resolves to the canonical lesson |
Attribution
Every response includes a structured attribution block:
{
"data": { ... },
"attribution": {
"cite_url": "https://blockchainacademics.com/...",
"as_of": "2026-04-19T12:34:56Z",
"source_hash": "sha256:..."
}
}
When your agent surfaces BCA content to a user, attribute via the cite_url. Fields are preserved as null when upstream omits them so downstream agents can detect missing provenance.
API Key
Not required to try it. When BCA_API_KEY is unset, the server falls back to a baked-in public demo key that unlocks 10 marquee tools. See Demo tier above.
For the full 105 tools, get a free key at brain.blockchainacademics.com/signup (2,000 calls/month). Paid tiers unlock agent-backed research generation and proprietary on-chain indicators.
Set BCA_API_KEY in your MCP client env. Optionally override BCA_API_BASE (default https://api.blockchainacademics.com). BCA_API_BASE_URL is also accepted for backward compatibility.
Errors
The server never crashes the stdio process. All failures surface as MCP responses with isError: true and a JSON body:
{ "error": { "code": "BCA_AUTH", "message": "..." } }
| Code | Meaning |
|---|---|
BCA_AUTH | Invalid BCA_API_KEY (HTTP 401/403) |
BCA_TIER_LOCKED | Tool not in your current tier's allowlist (HTTP 403). Demo tier sees this on 91 of the 105 tools — the error message includes the upgrade URL. |
BCA_RATE_LIMIT | Rate limit exceeded (HTTP 429 — honor Retry-After) |
BCA_UPSTREAM | BCA API returned 5xx or malformed JSON |
BCA_NETWORK | Network failure or 20s timeout exceeded |
BCA_BAD_REQUEST | Invalid tool arguments |
Development
npm install
npm run build # tsc -> dist/
npm test # node:test smoke suite
npm run dev # tsx src/index.ts (stdio)
Pre-commit hooks
This repo uses pre-commit to enforce formatting, lint, secret scanning, and large-file caps before every commit. Install once per clone:
pip install pre-commit # or: brew install pre-commit
pre-commit install # wires .git/hooks/pre-commit
pre-commit run --all-files # optional: lint the full tree now
Hooks configured in .pre-commit-config.yaml: prettier --check, eslint, gitleaks, detect-private-key, check-added-large-files (500KB cap).
Security
See SECURITY.md for vulnerability reporting, supported versions, and our 90-day coordinated disclosure policy.
License
MIT © 2026 Blockchain Academics
Comments
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.