Skip to content

Corporate credit data for AI agents — Python SDK, LangChain tools, and MCP server

License

Notifications You must be signed in to change notification settings

debtstack-ai/debtstack-python

Repository files navigation

DebtStack.ai Python SDK

Corporate credit data for AI agents.

PyPI License

Why DebtStack?

Equity data is everywhere. Credit data isn't.

There's no "Yahoo Finance for bonds." Corporate debt structures, guarantor chains, and covenant details are buried in SEC filings—scattered across 10-Ks, 8-Ks, credit agreements, and indentures. An AI agent trying to answer "which telecom companies have leverage above 5x?" would need to read dozens of filings, extract the right numbers, and compute ratios manually.

DebtStack fixes this. We extract, normalize, and serve corporate credit data through an API built for AI agents.

Three Things You Can't Do Elsewhere

1. Cross-Company Credit Queries

# Find distressed telecom companies
GET /v1/companies?sector=Telecommunications&min_leverage=5&sort=-net_leverage_ratio

Screen ~300 companies by leverage, coverage ratios, or maturity risk in one call. No filing-by-filing analysis.

2. Pre-Built Entity Relationships

# Who guarantees this bond?
POST /v1/entities/traverse
{"start": {"type": "bond", "id": "893830AK8"}, "relationships": ["guarantees"]}

Guarantor chains, parent-subsidiary hierarchies, structural subordination—mapped and queryable. This data exists nowhere else in machine-readable form.

3. Agent-Ready Speed

< 100ms response time

AI agents chain multiple calls. If each took 30 seconds (reading a filing), a portfolio analysis would take hours. DebtStack returns in milliseconds.


Installation

pip install debtstack-ai

For LangChain integration:

pip install debtstack-ai[langchain]

Quick Start

from debtstack import DebtStackClient
import asyncio

async def main():
    async with DebtStackClient(api_key="your-api-key") as client:

        # Screen for high-leverage companies
        risky = await client.search_companies(
            sector="Telecommunications",
            min_leverage=4.0,
            fields="ticker,name,net_leverage_ratio,interest_coverage",
            sort="-net_leverage_ratio"
        )

        # Drill into the riskiest one
        ticker = risky["data"][0]["ticker"]
        bonds = await client.search_bonds(ticker=ticker, has_pricing=True)

        # Check guarantor coverage on their notes
        for bond in bonds["data"]:
            guarantors = await client.get_guarantors(bond["cusip"])
            print(f"{bond['name']}: {len(guarantors)} guarantors")

asyncio.run(main())

Synchronous Usage

from debtstack import DebtStackSyncClient

client = DebtStackSyncClient(api_key="your-api-key")
result = client.search_companies(sector="Energy", min_leverage=3.0)

What's In The Data

Coverage Count
Companies ~300 (S&P 100 + NASDAQ 100 + high-yield issuers)
Entities ~39,000 (subsidiaries, holdcos, JVs, VIEs)
Debt Instruments ~10,000 (bonds, loans, revolvers) with 96% document linkage
Bond Pricing ~4,300 bonds with FINRA TRACE pricing (updated 3x daily)
SEC Filing Sections ~25,000 (searchable full-text)
Covenants ~1,800 structured covenant records

Pre-computed metrics: Leverage ratios, interest coverage, maturity profiles, structural subordination scores.

Relationships: Guarantor chains, issuer-entity links, parent-subsidiary hierarchies.

API Methods

Method What It Does
search_companies() Screen by leverage, sector, coverage, risk flags
search_bonds() Filter by yield, spread, seniority, maturity
resolve_bond() Look up CUSIP, ISIN, or "RIG 8% 2027"
traverse_entities() Follow guarantor chains, map corporate structure
search_pricing() FINRA TRACE bond prices, YTM, spreads
search_documents() Full-text search across credit agreements, indentures
batch() Run multiple queries in parallel
get_changes() Track debt structure changes over time

Examples

Which MAG7 company has the most debt?

result = await client.search_companies(
    ticker="AAPL,MSFT,GOOGL,AMZN,NVDA,META,TSLA",
    fields="ticker,name,total_debt,net_leverage_ratio",
    sort="-total_debt",
    limit=1
)
# Returns structured data in milliseconds, not minutes

Find high-yield bonds trading at a discount

result = await client.search_bonds(
    seniority="senior_unsecured",
    min_ytm=8.0,
    has_pricing=True,
    sort="-pricing.ytm"
)

Who guarantees a specific bond?

guarantors = await client.get_guarantors("893830AK8")
for g in guarantors:
    print(f"{g['name']} ({g['entity_type']}) - {g['jurisdiction']}")

# Output:
# Transocean Ltd. (holdco) - Switzerland
# Transocean Inc. (finco) - Cayman Islands
# Transocean Offshore Deepwater Drilling Inc. (opco) - Delaware
# ... 42 more entities

Search for covenant language

result = await client.search_documents(
    q="maintenance covenant",
    section_type="credit_agreement",
    ticker="CHTR"
)
# Returns matching sections with highlighted snippets

LangChain Integration

Build AI agents that can autonomously analyze corporate credit data.

Installation

pip install debtstack-ai[langchain]

Available Tools

Tool Description
debtstack_search_companies Screen companies by leverage, sector, coverage ratios, and risk flags
debtstack_search_bonds Filter bonds by yield, spread, seniority, maturity, and pricing
debtstack_resolve_bond Look up bonds by CUSIP, ISIN, or description (e.g., "RIG 8% 2027")
debtstack_traverse_entities Follow guarantor chains, map corporate structure, trace ownership
debtstack_search_pricing Get FINRA TRACE bond prices, YTM, and spreads
debtstack_search_documents Full-text search across credit agreements, indentures, and SEC filings
debtstack_get_changes Track debt structure changes over time (new issuances, maturities, leverage)

Full Example

from debtstack.langchain import DebtStackToolkit
from langchain.agents import AgentExecutor, create_openai_functions_agent
from langchain_openai import ChatOpenAI
from langchain import hub

# Initialize toolkit with your API key
toolkit = DebtStackToolkit(api_key="your-api-key")
tools = toolkit.get_tools()

# Create an agent with GPT-4 (or any LangChain-compatible LLM)
llm = ChatOpenAI(temperature=0, model="gpt-4")
prompt = hub.pull("hwchase17/openai-functions-agent")
agent = create_openai_functions_agent(llm, tools, prompt)
agent_executor = AgentExecutor(agent=agent, tools=tools, verbose=True)

# Ask natural language questions about corporate credit
result = agent_executor.invoke({
    "input": "Which telecom companies are most at risk of default?"
})
print(result["output"])

Example Queries

The agent can handle complex, multi-step credit analysis:

  • "Which telecom companies have leverage above 5x and near-term maturities?"
  • "Find all bonds yielding above 8% with senior secured status"
  • "Who guarantees Transocean's 8.75% 2030 notes? How many entities are in the guarantee chain?"
  • "Compare Charter and Altice's corporate structures - which has more structural subordination risk?"
  • "What changed in RIG's debt structure since January 2025?"
  • "Search for maintenance covenant language in Charter's credit agreements"
  • "Find distressed bonds trading below 80 cents on the dollar"

MCP Server (Claude Desktop, Claude Code, Cursor)

Give Claude (or any MCP client) direct access to corporate credit data.

Installation

pip install debtstack-ai[mcp]

Available Tools

Tool Description
search_companies Search companies by ticker, sector, leverage ratio, and risk flags
search_bonds Search bonds by ticker, seniority, yield, spread, and maturity
resolve_bond Look up a bond by CUSIP, ISIN, or description (e.g., "RIG 8% 2027")
get_guarantors Find all entities that guarantee a bond
get_corporate_structure Get full parent-subsidiary hierarchy for a company
search_pricing Get FINRA TRACE bond prices, YTM, and spreads
search_documents Search SEC filing sections (debt footnotes, credit agreements, indentures)
get_changes See what changed in a company's debt structure since a date

Claude Desktop

Add to your Claude Desktop config (~/.config/claude/mcp.json on Mac/Linux, %APPDATA%\Claude\mcp.json on Windows):

{
    "mcpServers": {
        "debtstack-ai": {
            "command": "debtstack-mcp",
            "env": {
                "DEBTSTACK_API_KEY": "your-api-key"
            }
        }
    }
}

Claude Code

Add to your Claude Code config (~/.claude/mcp.json):

{
    "mcpServers": {
        "debtstack-ai": {
            "command": "debtstack-mcp",
            "env": {
                "DEBTSTACK_API_KEY": "your-api-key"
            }
        }
    }
}

Cursor

Add to your Cursor MCP settings (.cursor/mcp.json):

{
    "mcpServers": {
        "debtstack-ai": {
            "command": "debtstack-mcp",
            "env": {
                "DEBTSTACK_API_KEY": "your-api-key"
            }
        }
    }
}

Alternative: Run with Python Module

If you prefer not to install the console script, you can use python -m instead:

{
    "mcpServers": {
        "debtstack-ai": {
            "command": "python",
            "args": ["-m", "debtstack.mcp_server"],
            "env": {
                "DEBTSTACK_API_KEY": "your-api-key"
            }
        }
    }
}

Example Queries

Once configured, ask Claude:

  • "Which energy companies have near-term maturities and high leverage?"
  • "Who guarantees the Transocean 8% 2027 notes?"
  • "Compare Charter's debt structure to Altice"
  • "Find all senior secured bonds yielding above 8%"
  • "What are the financial covenants in Charter's credit agreement?"
  • "Show me RIG's corporate structure - where does the debt sit?"
  • "What changed in CVS's debt structure since June 2025?"
  • "Search for change-of-control provisions in Altice's indentures"

Pricing

DebtStack offers usage-based pricing with a free tier to get started.

See debtstack.ai/pricing for details.

Links

License

Apache-2.0

About

Corporate credit data for AI agents — Python SDK, LangChain tools, and MCP server

Topics

Resources

License

Stars

Watchers

Forks

Packages

 
 
 

Contributors