Toolstem Sec Mcp Server

SEC EDGAR MCP Server — Insider Signals, 13D Activist Risk & Filing Intelligence

[npm version] [npm downloads] [License: MIT]

SEC EDGAR intelligence for AI agents. Five composite tools that pre-compute high-value signals directly from SEC EDGAR's public submissions API, returned as structured JSON.

No SEC API key required. Data is sourced directly from SEC EDGAR's public submissions API. A built-in sliding-window rate limiter keeps traffic under SEC's 10 rps fair-access ceiling automatically.

Quickstart — hosted endpoint (recommended)

Point your MCP client or agent at the hosted endpoint. No API key, no infra, no setup. Billing is per-call via x402 — the agent's wallet pays directly in USDC on Base mainnet.

https://mcp.toolstem.com/mcp/sec

• No SEC API key, no signup, no marketplace account — the agent's wallet pays directly.
• No infrastructure — nothing to install, host, or keep running.
• No setup — connect an MCP client and call a tool.
• initialize and tools/list are free (discovery and schema introspection).
• tools/call is tiered per tool (see Pricing below).

Claude Desktop

Drop this into your claude_desktop_config.json:

{  "mcpServers": {    "toolstem-sec": {      "command": "npx",      "args": [        "-y",        "mcp-remote",        "https://mcp.toolstem.com/mcp/sec"      ]    }  }}

Restart Claude Desktop, then ask: "Has TSLA disclosed any material 8-K events in the last 90 days?"

Any MCP client (LangChain.js)

The official @langchain/mcp-adapters library connects directly to the hosted URL:

import { MultiServerMCPClient } from "@langchain/mcp-adapters";import { ChatOpenAI } from "@langchain/openai";import { createReactAgent } from "@langchain/langgraph/prebuilt";
const client = new MultiServerMCPClient({  toolstem_sec: {    transport: "http",    url: "https://mcp.toolstem.com/mcp/sec",    // Add your x402-signing middleware via headers, OR run an x402    // proxy locally and point url at it. See https://www.x402.org/clients.  },});
const tools = await client.getTools();const agent = createReactAgent({ llm: new ChatOpenAI({ model: "gpt-4o-mini" }), tools });await agent.invoke({ messages: "Has TSLA disclosed any material 8-K events in the last 90 days?" });

LangChain quick-start (langchain-toolstem)

The langchain-toolstem wrapper handles x402 payment for you — pass a funded wallet key and the SEC tools are included automatically:

import { createToolstemTools } from 'langchain-toolstem';const tools = await createToolstemTools({ walletPrivateKey: process.env.WALLET_KEY });// SEC tools included automatically — agents pay per call in USDC

Prefer to run the server yourself over stdio/HTTP? See Advanced: self-host at the bottom.

Try the tools live in the Toolstem playground.

Pricing

• MCP initialize and tools/list are free — discover the server and its tool surface without paying anything.
• tools/call is tiered per tool, paid in USDC on Base mainnet via x402. No API key, no signup, no marketplace account — agents pay directly from their own wallet.

Per-tool breakdown:

How billing works

Toolstem uses the x402 payment protocol. Agents pay per call in USDC on Base — no API keys, no subscriptions, no invoices. The agent's wallet settles each call automatically via EIP-3009.

See the live pricing page on toolstem.com/sec/ for current rates.

The five tools

All five tools are composite/curated (they compute derived signals or aggregate across multiple EDGAR endpoints — no raw passthroughs). Annotations: readOnlyHint: true, destructiveHint: false, idempotentHint: true, openWorldHint: true.

1. get_company_filings_summary

Overview of a company's filing activity: last 20 filings + computed signals.

Example output (abbreviated):

{  "ticker": "AAPL",  "cik": "0000320193",  "company_name": "Apple Inc.",  "signals": {    "filing_velocity": "NORMAL",    "material_event_count_90d": 4,    "disclosure_volume_trend": "RISING",    "latest_form_types": ["8-K", "4", "DEF 14A"]  },  "meta": { "source": "sec_edgar_direct", "data_delay": "live" }}

2. get_insider_signal

Probes Form 3 / 4 / 4/A insider filing activity within a configurable lookback window. Required: ticker_or_cik. Optional: lookback_days (1–730, default 90).

Returns: recent_insider_filings[] (accession numbers + SEC URLs), net_transaction_count, buy_count, sell_count, and insider_signal.

v0.1 limitation — counts only. v0.1 returns counts and Form 4 references only; direction-aware buy/sell signals ship in v0.2 (Form 4 XML parsing). Today, insider_signal is null when filings exist in the window (direction unknown) and "NEUTRAL" when no insider filings exist (verified absence). buy_count / sell_count are 0 in v0.1.

Example output (abbreviated):

{  "ticker": "MSFT",  "cik": "0000789019",  "company_name": "MICROSOFT CORP",  "lookback_days": 90,  "insider_signal": null,  "net_transaction_count": 0,  "buy_count": 0,  "sell_count": 0,  "recent_insider_filings": [    {      "accession_number": "0001127602-26-001234",      "filing_date": "2026-04-15",      "sec_url": "https://www.sec.gov/Archives/edgar/data/789019/000112760226001234/0001127602-26-001234-index.htm"    }  ],  "meta": { "source": "sec_edgar_direct", "data_delay": "live" }}

3. get_institutional_signal

Probes for activist investor activity via SC 13D / 13D/A filings. Required: ticker_or_cik. Optional: quarters_back (1–20, default 4 ≈ 1 year).

v0.1 limitation — activist flag only. v0.1 ships the live activist_risk_flag (from 13D/13D-A) and a list of 13D filings. Quarterly 13F XBRL parsing — which produces institutional_signal (ACCUMULATING / HOLDING / DISTRIBUTING) and recent_13f_count — ships in v0.2. Today those two fields are null / 0.

Example output (abbreviated):

{  "ticker": "NVDA",  "cik": "0001045810",  "company_name": "NVIDIA CORP",  "quarters_back": 4,  "institutional_signal": null,  "recent_13f_count": 0,  "activist_risk_flag": false,  "recent_13d_filings": [],  "meta": { "source": "sec_edgar_direct", "data_delay": "live" }}

4. get_material_events_digest ⚡ Premium tier

Premium tier — $0.50 USDC per call on Base mainnet, settled via x402. See the live pricing page on toolstem.com/sec/ for current rates.

Severity-ranked digest of all 8-K and 8-K/A filings within a configurable lookback window. Each item code is mapped to a plain-English label and severity rating. Required: ticker_or_cik. Optional: lookback_days (1–1825, default 365).

Returns: events[] (sorted newest-first), redflag_count, category_counts.

Example output (abbreviated):

{  "ticker": "TSLA",  "cik": "0001318605",  "company_name": "Tesla, Inc.",  "lookback_days": 180,  "redflag_count": 1,  "category_counts": { "RED": 1, "YELLOW": 3, "GREEN": 7 },  "events": [    {      "accession_number": "0001628280-26-005678",      "filing_date": "2026-04-10",      "form": "8-K",      "items": [        { "code": "4.02", "label": "Non-Reliance on Previously Issued Financial Statements", "category": "financial", "severity": "RED" }      ],      "sec_url": "https://www.sec.gov/Archives/edgar/data/1318605/000162828026005678/0001628280-26-005678-index.htm"    }  ],  "meta": { "source": "sec_edgar_direct", "data_delay": "live" }}

5. compare_disclosure_signals

Side-by-side comparison of 2–5 companies across all key disclosure signals. Required: tickers_or_ciks (string[2..5]). All lookups run in parallel.

Returns per-company: filing_velocity, material_event_count_90d, redflag_count_365d, activist_risk_flag, last_filing_date.

Returns winners (as CIKs, not tickers — cross-reference with the companies[] array): quietest_disclosure, most_active, most_redflags, activist_targets.

Example output (abbreviated):

{  "companies": [    {      "ticker": "AAPL",      "cik": "0000320193",      "filing_velocity": "NORMAL",      "material_event_count_90d": 4,      "redflag_count_365d": 0,      "activist_risk_flag": false,      "last_filing_date": "2026-04-25"    },    {      "ticker": "MSFT",      "cik": "0000789019",      "filing_velocity": "ACCELERATING",      "material_event_count_90d": 7,      "redflag_count_365d": 0,      "activist_risk_flag": false,      "last_filing_date": "2026-04-26"    }  ],  "winners": {    "quietest_disclosure": "0000320193",    "most_active": "0000789019",    "most_redflags": null,    "activist_targets": []  },  "meta": { "source": "sec_edgar_direct", "data_delay": "live" }}

Advanced: self-host

Most users should use the hosted endpoint above — it needs no API key, no infrastructure, and no setup. This section is for users who specifically want to run the server themselves. When self-hosting you are responsible for running the process and for supplying an EDGAR fair-access contact (SEC_USER_AGENT_CONTACT).

Claude Desktop (self-hosted over stdio)

Run locally over stdio — no x402 charges, you bring your own EDGAR fair-access contact:

{  "mcpServers": {    "toolstem-sec": {      "command": "npx",      "args": ["-y", "toolstem-sec-mcp-server"],      "env": {        "SEC_USER_AGENT_CONTACT": "[email protected]"      }    }  }}

npm (MCP stdio transport)

npm install -g toolstem-sec-mcp-servertoolstem-sec-mcp-server

Self-hosted HTTP

Three modes:

Local-only (default — safest):

toolstem-sec-mcp-server --http# Binds 127.0.0.1:3000 — reachable only from this machine

Remote with auth:

ALLOW_REMOTE=1 MCP_AUTH_TOKEN=my-secret toolstem-sec-mcp-server --http# Binds 0.0.0.0:3000 — requires Bearer token on every /mcp request

Remote without auth (use at your own risk):

ALLOW_REMOTE=1 MCP_AUTH_DISABLED=1 toolstem-sec-mcp-server --http# Binds 0.0.0.0:3000 — no authentication

SEC EDGAR fair-access policy

All outbound traffic goes through a shared sliding-window rate limiter (8 rps target, 4 rps safety margin below SEC's 10 rps hard cap). Every request includes a User-Agent header identifying the package and a contact email per SEC policy. Override the contact email via:

[email protected] toolstem-sec-mcp-server

Violating SEC's fair-access policy can result in your IP being blocked. This server is designed to stay compliant automatically.

v0.2 roadmap

• Form 4 XML parsing — direction-aware insider signals (STRONG_BUYING / BUYING / NEUTRAL / SELLING / STRONG_SELLING) with net share counts
• 13F XBRL parsing — quarterly institutional flow signals (ACCUMULATING / HOLDING / DISTRIBUTING) with institution count
• 8-K text extraction — natural-language summaries of each material event from the filing's primary HTML document

License & author

MIT License — see LICENSE.

Built by Toolstem. Data sourced directly from SEC EDGAR.