Context API — AI-Ready Financial Context in One Call

Q: Which mode should I start with?

lite if you already have your own LLM/agent and just want sourced data. standard if you want a ready-to-use written answer. advanced only if you need the extra thoroughness of a gap-fill retry and can accept the higher cost and latency.

The Context API answers a natural-language financial question in a single call. Send a question like “Bitcoin sentiment over the last week” or “how is the ECB rate decision affecting EUR/USD” and get back exactly the sources you need — news, sentiment, prices, and our own analysis — either as raw structured data for your own agent to reason over, or as a ready-to-use written answer with cited sources.

Try it in the playground FAQ

Endpoint: POST /api/v1/context
Auth: same API key as the rest of the SentiSignal API (Authorization or x-api-key header).

Quick Start

mode=lite — cheapest, raw structured data, no synthesis

curl -X POST "https://www.sentisignal.live/api/v1/context" \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"query": "What is driving Bitcoin sentiment this week?", "mode": "lite"}'

mode=standard — lite + a written, anti-plagiarism-checked answer

curl -X POST "https://www.sentisignal.live/api/v1/context" \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"query": "How is the ECB rate decision affecting EUR/USD?", "mode": "standard"}'

mode=advanced — standard + a bounded gap-fill loop for thin first passes

curl -X POST "https://www.sentisignal.live/api/v1/context" \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"query": "What is driving Bitcoin sentiment this week?", "mode": "advanced"}'

Request body parameters

Field	Type	Required	Description
query	string	required	Your natural-language financial question, up to 2000 characters
mode	string	optional	`lite` (default) \| `standard` \| `advanced`
assetType	string	optional	Hint the asset class when your question is ambiguous — `crypto` \| `commodity` \| `fx`
symbol	string	optional	Pin a specific symbol instead of relying on auto-detection from `query`
limit	integer	optional	1–1000. Note: regardless of this value, at most 10 news items ever appear in the response — see “Limits” below.

Three modes available today

Lite mode (mode=lite, the default)

Raw, structured data, no prose. A fast routing step (one LLM call) picks the relevant sources for your question and returns them as-is: news items with AI-generated summaries, sentiment scores, licensed price data, and more. Nothing is written or paraphrased for you — this mode is built for callers who want to feed the result into their own LLM or agent and write their own answer. It is also the cheapest and fastest mode, and it is what you get if you omit the mode parameter entirely, so a quick integration never accidentally triggers a more expensive call.

Standard mode (mode=standard)

Everything lite mode returns, plus a ready-to-read written answer (answer field) with the sources it drew on (sources field). We run this answer through an anti-plagiarism safeguard before returning it, so you never receive verbatim copied text from any source — only an original paraphrase in our own words, grounded in and citing the real sources behind it.

Advanced mode (mode=advanced)

Everything standard mode does, plus a bounded gap-fill loop: before writing the final answer, a fast assessor model checks whether the first pass found enough to answer well, and if not, runs a further bounded round of retrieval to fill the gap. Most requests finish in exactly the same shape and cost as standard mode — the extra round only fires when the first pass genuinely looks thin, and the loop is capped so a single request can never run away. This is the most thorough mode and the most expensive of the three.

What you will never get back from any mode

Cryptocurrency prices. Our crypto price data is licensed for internal use only, not for redistribution through an API, so it is used behind the scenes to ground routing and never appears in your response. If your question resolves only to a crypto price with nothing else to show, you will get a clear “no sources found” response rather than a raw price.
Full article text. News results always come back as an AI-generated summary plus a link to the original publisher, never the publisher’s full original text — this protects publisher copyright and keeps every response concise.
Our own content (articles we publish on SentiSignal) is the one exception — since we own it, it comes back in full.

Limits and credits

At most 10 news items per call, regardless of how broad your question is or what limit you request. This keeps every response focused and fast to read.
Price data comes back at a granularity that adapts to your question’s time window — a one-day question gets fine granularity, a one-year question gets coarser data, automatically.
Context API shares one monthly credit pool with News Search API (both are “content” products). Each call costs credits proportional to how much it actually runs: lite = 2 credits, standard = 6 credits, advanced = 10 credits — advanced costs more because it can run a further LLM round to fill gaps. Every response includes X-RateLimit-Pool and X-RateLimit-Credits-Charged headers so you always know what a call cost.
The functional surface — which modes, filters, and data types are available — is identical on every paid plan. Plans differ only in how many credits you get per period, never in what you can access. The one exception is the crypto-price restriction above, which applies equally to every plan because it is a data-licensing rule, not a plan feature.

Error handling

Every response is either a successful result or a clear, structured error — an invalid request, a missing or invalid API key, a 429 when your credit pool is exhausted, a 404 “no sources found” for an unanswerable question, or (standard/advanced only) a 422 for the rare case where we could not produce an answer we were confident was sufficiently original, in which case we ask you to narrow your question rather than return something we are not confident about.

FAQ

How is Context API different from News Search API?

News Search returns a ranked list of matching articles for you to read or process yourself. Context API answers a question directly — it searches the same corpus, then (in standard/advanced mode) writes one synthesized, cited answer. Use News Search when you want results; use Context when you want an answer.

Which mode should I start with?

lite if you already have your own LLM/agent and just want sourced data. standard if you want a ready-to-use written answer. advanced only if you need the extra thoroughness of a gap-fill retry and can accept the higher cost and latency.

Does Context API count against the same limit as News Search?

Yes — both share one “content” credit pool, separate from the Metals/FX/Commodity “price” pool. See “Limits and credits” above for exact weights per mode.

Can I get raw Bitcoin/crypto prices from this endpoint?

No. Crypto price data is licensed for internal use only and is never returned by Context API, even though it may be used behind the scenes to route your question. This applies to every plan.

Can I test this before writing code?

Yes — the API playground lets you run real requests against your own key (or a free-tier key) directly from the browser, for all three modes.

Get started

Use your existing SentiSignal API key, or try it first in the playground. See the full technical reference for exact request/response shapes and error codes.