OMNI Datastream vs edgartools

Feature-by-feature comparison of OMNI Datastream and edgartools — speed, cost, agent support, and commercial posture

Preliminary — pending external methodology review (OMNI-3064). Numbers on this page will be re-verified by an external SRE before this comparison is linked from the docs sidebar. See benchmarks/METHODOLOGY.md.

OMNI Datastream vs edgartools

edgartools is a well-designed open-source Python library (MIT-licensed) that talks directly to SEC EDGAR. It’s the OSS peer most agent-builders evaluate alongside hosted APIs. This comparison puts OMNI head-to-head against edgartools on the same four workflows we use to benchmark against sec-api.io and financialdatasets.ai.

Speed

From a ubuntu-latest GitHub Actions runner, N=5 iterations per case. Every edgartools iteration runs in a fresh subprocess with all caches cleared, so each measurement is a truly cold SEC fetch — no in-memory or on-disk cache hits skew the p50 (methodology):

Operation	OMNI p50	edgartools p50	Speedup	OMNI wire bytes	edgartools wire bytes
Entity resolve	34ms	471ms	13.9×	273	164,686
Filing search	38ms	688ms	18.1×	1,033	362,459
Section extract	34ms	3,081ms	90.6×	2,954	3,053,868
XBRL → JSON	34ms	1,508ms	44.3×	1,522	1,520,790

edgartools fetches raw SEC documents client-side (e.g., ~3 MB of 10-K HTML to extract Item 1A text) before parsing. That’s an architectural difference — edgartools optimizes for pip install, zero API key, SEC direct and accepts the bandwidth cost. OMNI pre-parses once in its cluster and returns compact JSON.

Pricing

Tier	OMNI	edgartools
Free	250 calls/month (renewable)	Unlimited (truly free)
Pay-as-you-go	From $0.01/call	Not applicable
Personal	$55/month	$0
Business	$239/month	$0
Enterprise	Custom (SLA, support, dedicated capacity)	None (community-supported)

The honest take: edgartools is genuinely free. If you can tolerate client-side parsing of multi-megabyte SEC documents, aren’t running at agent scale where token cost or latency matters, and don’t need an SLA — it’s a great option. OMNI costs money; the value is operational:

• Latency 13.9-90.6× faster on the same queries
• Payloads 600-1,000× smaller (fewer LLM tokens)
• SLA, support, dedicated capacity
• Commercial redistribution rights
• Hosted semantic search, dilution intelligence, MCP server, monitors, etc.
• Native JS/TS + Go + Rust SDKs (edgartools is Python-only)

Where edgartools wins

edgartools has genuine strengths. Customers evaluating both should know:

• $0 cost, MIT license — no procurement, no monthly floor.
• Code you can read — the source is on GitHub; auditable and modifiable.
• No third-party dependency in your stack — edgartools → SEC direct.
• Clean, agent-friendly Python API — well-typed, thoughtful ergonomics.
• Full-text search & XBRL parsing — all client-side, no vendor coupling.
• Permissive redistribution for the library itself (MIT).

If your SEC-data workload is research-scale (one ticker at a time, small number of filings), edgartools is often the right choice. OMNI’s value compounds with scale, agents, and production requirements.

Where OMNI wins

Dimension	OMNI	edgartools
Latency (hot path)	34-38ms p50	300-2,100ms p50
Payload size (agent-ingestable)	Compact JSON	Raw 10-K HTML / full XBRL
Semantic search	Yes (Pinecone + Voyage hybrid)	No
Intelligence bundles	Yes (company, security, earnings)	No (build it yourself)
Filing diff / amendment tracking	Yes	No
Monitors + webhooks	Yes (signed, replay, cursor)	No
MCP native server	Yes	No
SDKs	JS, Python, Go, Rust	Python-only
CLI (omni-sec on npm)	Yes	No (Python import only)
Status page / SLA	Yes	No (best-effort OSS)
Commercial redistribution license	Yes (via commercial agreement)	No (library itself: MIT; SEC data: public)
8-K 5.07 voting results (structured)	Yes	No
Filing export formats	JSON, MD, CSV, XLSX, DOCX, PDF	Python objects only
13F comparison (hedge-fund overlap)	Yes	No (manual aggregation)
Factor analysis	Yes (84 factors)	No
Macro data	Yes (12 countries)	No (SEC-only)

When to choose which

Choose edgartools when:

• Cost is the hard constraint ($0 matters).
• Workload is small (hundreds of filings/year, not millions).
• You’re building a research script or one-off notebook, not a production agent.
• You want to read/modify the library source.
• You’re comfortable parsing raw SEC documents client-side.

Choose OMNI Datastream when:

• You’re running an agent workload (latency + token cost compound).
• You need SLA, support, or commercial redistribution.
• You want semantic search, monitors, intelligence bundles, or MCP out of the box.
• You’re on a non-Python stack (JS, Go, Rust).
• You need 8-K 5.07 voting results, filing diffs, or other structured data edgartools doesn’t extract.

Use both if you’re prototyping in a notebook (edgartools) and shipping in production (OMNI) — that’s a common pattern.

Migration sketch

# edgartools → OMNI

# 1. Entity resolve
# edgartools
from edgar import Company, set_identity
set_identity("You [e-mail]")
c = Company("AAPL")
cik, name = c.cik, c.name

# OMNI Datastream (JS/TS)
const entity = await omni.entities.resolve({ ticker: "AAPL" });
// returns { cik, name, ticker, ... }

# 2. Latest 10-K
# edgartools
filing = c.get_filings(form="10-K").latest(1)

# OMNI
const latest = await omni.filings.latest({ ticker: "AAPL", form: "10-K" });

# 3. Item 1A
# edgartools (pulls full 10-K HTML client-side)
item_1a = filing.obj().risk_factors

# OMNI
const item1A = await omni.filings.sections.latest({
  ticker: "AAPL", form: "10-K", item: "item_1a"
});

# 4. Balance sheet from XBRL
# edgartools
bs = filing.xbrl().statements.balance_sheet()

# OMNI
const balanceSheet = await omni.statements.balance_sheet({
  ticker: "AAPL", period: "annual", limit: 2
});

Reproducibility

All numbers on this page come from public scripts:

• scripts/bench/benchmark_search.py (OMNI)
• scripts/bench/benchmark_edgartools.py (edgartools)
• scripts/bench/competitive_scorecard.py (4-way aggregator)
• benchmarks/METHODOLOGY.md (reviewable methodology doc)

Run them yourself. If you find a methodology flaw, file an issue or open a PR.