DEVELOPER REFERENCE / V1

Docs

Statiq returns reviewed events with their status, confidence and source receipts, through a REST API and an MCP server. Responses are JSON (conflict events can also come as CSV). Status and confidence are returned exactly as recorded.

Authentication

Send your API key as a bearer token with every request.

  • Sign up at /signup with your name and email, or with Google. New accounts join a waitlist; API keys are available once Statiq lets the account in.
  • Create your key under Console → API keys, or on the setup screen. Each account holds one key of its own, shown once. Statiq can also issue keys.
  • Keys are scoped and stored hashed. Revoke a key in the console. Never put a key in a URL.
Base URL: https://statiq.sh/api/public/v1
Authorization: Bearer $STATIQ_API_KEY
Content-Type: application/json

The hosted base path is /api/public/v1, not a bare /v1. A key uses its account’s plan; a key Statiq issues is capped at its own tier. The monthly allowance counts every key on the account, REST and MCP together, and resets at 00:00 UTC on the 1st. The per-minute rate applies to each key.

Plans and limits

PlanPriceCalls a monthA minuteDelayHistoryMCPPlacesChatInvestigations a dayText alerts a day
IndividualFree2001060 min30 daysread-only, 20 a day, 10 results1 · home area4 a week1none
Professional trial (7 days)free, no card1,00060none12 monthsevery tool1020 / 5 h, 30 a day33
Professional$15/month1,00060none12 monthsevery tool1020 / 5 h, 30 a day33
Complete$56/month5,000120none12 monthsevery tool50100 / 5 h, 300 a day3010
Analyst (Finance)$149/month1,00060none12 monthsevery tool1020 / 5 h, 30 a day33
Desk (Finance)$899/month25,000120none12 monthsevery tool50100 / 5 h, 300 a day3010

Individual is free for personal use, evaluation and building against the API and MCP. Use by or for an organisation needs Professional or Complete. Statiq invoices Professional ($15/month) and Complete ($56/month) during the alpha. Each account gets one 7-day Professional trial with Professional’s limits and no card, starting when setup is finished, unless Individual is chosen at sign-up. Volumes beyond Complete are agreed with Statiq.

Statiq Finance (pre-alpha, by invitation) has two plans that include a Statiq plan, so an account pays one subscription: Analyst ($149/month) includes Professional with its limits, and Desk ($899/month) includes Complete with 25,000 API and MCP calls a month. An account holds one plan, so the plan field in answers and refusals can be finance_analyst or finance_desk.

  • Delay. On Individual, reports and alerts arrive 60 minutes after detection. Official life-safety warnings, such as severe-weather warnings and evacuation orders, are real time on every plan.
  • History. Lists reach back 30 days on Individual and 12 months on Professional and Complete. An earlier since is moved to the start of your history, and the answer gives history_from.
  • Places. Places count the home area: Individual watches its home area only (1), Professional 10 places and Complete 50. Places beyond the plan stay saved but paused (no alerts, not searched) until the plan covers them.
  • Alerts. Urgent alerts go by email, Telegram and your webhook, only for confirmed or corroborated reports; official life-safety warnings go at once. Other official records, such as residents’ 311 requests, city data and road closures, alert only once a second independent source backs them.
  • Text messages (not available yet): none on Individual, 3 a UTC day on Professional and its trial, 10 on Complete.
  • Priority. On Complete, your places are collected and reviewed first, within each source’s own limits. No latency is promised.

Rate limits

Every REST response carries your remaining allowance in headers:

HeaderCounts
X-RateLimit-Limit, X-RateLimit-Remaining, X-RateLimit-ResetThe month; Reset is in epoch seconds.
X-RateLimit-Minute-Limit, X-RateLimit-Minute-RemainingThis key’s current minute.
X-RateLimit-Day-Limit, X-RateLimit-Day-RemainingIndividual’s MCP calls in the UTC day.

Over MCP, a single tools/call request (not a batch) answers with the same headers, and the tool result’s structuredContent.usage carries the same counts. A refused call answers 429 with Retry-After (seconds to the reset) and a JSON body naming the limit, the count used and the reset time:

{"error":"Monthly API call allowance reached","plan":"individual","limit":200,"used":200,"resets_at":"2026-12-01T00:00:00Z"}

Errors

Errors are JSON with an error field.

StatusMeaning
400Invalid request.
401Missing or revoked key.
403Missing scope, or a plan without access (conflict data needs Professional or Complete).
404Unknown source, story or conflict event, or one your key cannot see.
413A verify_claim body over 10,000 bytes.
429Allowance or rate limit reached; see Retry-After.
503Service unavailable.

Transparency & credibility tracking

Statiq keeps three things apart: the evidence behind a report, its review state and the credibility of its sources. A reviewed report is not a confirmed claim, and a source score does not guarantee that any report is correct.

Evidence you can inspect

Every event response includes the report’s status and confidence and, for each public source, its ID, original link and posting time.

  • A source link identifies the supporting report, not necessarily a primary record. Only public sources are cited; internal intake and review records are never returned.
  • posted_at is when the source published; first_detected_at is when Statiq detected the report. Neither says when a claim was verified.
  • Summaries are reading aids, not replacements for the cited original.
  • corrections lists each correction in force with its date and what changed (for a headline or summary, the earlier wording). A correction never changes a report’s status or claim history.
  • Reports are audited against Statiq’s editorial standards, and signed-in readers can report an error on any report they can see.

What reaches the event feed

Events come from local reporting and official records as well as national and world outlets. Each kind is worded for what it is:

Local news
Local and regional outlets’ reports of incidents happening now, reviewed by the same live-event rule as every other report.
Official local records
Storm and weather reports, road closures and incidents, and transit service alerts, each at the record’s own coordinates where it has them. Transit alerts are live disruptions only: suspensions, and cuts, delays, detours or moved stops caused by an accident, weather, a strike, a demonstration or police activity, or marked severe by the agency. Planned work, holiday timetables and elevator or accessibility notices are left out.
Residents’ requests
City 311 and similar service requests are worded as requests (“a 311 request reports…”), not as confirmation of an incident. Reporters’ names and contact details are never kept.
Police dispatch records
Published calls for service from a few cities, kept only for serious incidents that affect the area: shootings and shots fired, stabbings, explosions, bomb threats and suspicious devices, barricade or hostage situations, major police operations and road closures, hazardous materials, fires, major crashes, and aircraft or train incidents. Each is worded as a call for service (“not a confirmed crime, and call types can change”), placed no closer than its block or intersection, and never treated as an official record. Arrests, warrants, missing persons, medical, mental-health and welfare calls, and crimes against people or property are left out; no names, notes or officers are read.
Worldwide official records
Orange and red disaster alerts, earthquakes of magnitude 5 and above, and satellite-observed wildfires, volcanic activity, severe storms, floods and landslides (planned burns are left out).
Internet outages
A country or region whose measured connectivity falls below half of normal: a measurement, not a report of the cause. Outage measurements appear in the console only. The API and MCP never list them as sources and leave out a report that rests on them alone.
Grid status
A status California’s grid operator declares (a Flex Alert, restricted maintenance or an energy emergency, and the return to normal) is an official record.
Other languages
Reports in other languages are reviewed by the same rule. Titles and summaries are in English; the source link points to the original-language report, and the place is taken from how the report itself names it.

Copies of one agency story count as one source, and repeated posts or shares never corroborate themselves. A live event is never left out for being single-source, brief or unverified, and no kind above confirms a claim on its own.

Article review vs. claim confirmation

A report is published after a recorded AI review or an audited human confirmation. AI review checks that the report is faithful to its source, attributed and categorized, and placed only where the reporting supports. It can publish a single-source report whose claim stays unverified.

Unverified
Not independently confirmed, even when the report has passed review.
Corroborated
At least two independent sources support it; a weather-station or stream-gauge reading can be one of them. This is not the final confirmation standard: check the citations and their independence.
Confirmed
Settled by the settlement squad or by an administrator’s attested review, and settled_by says which. The squad confirms a corroborated report at least 30 minutes old, or 10 minutes old when two official agencies agree, when three or more independent sources, or two including an official one, agree in verbatim quotes and no correction contradicts it. The 10-minute path closes by itself while more than 2% of its confirmations are undone. A confirmation made in a report’s first six hours is reopened if a correction appears within a day. An administrator can undo a squad settlement.
False
A recorded false-claim outcome. The squad settles false only after six hours, when an official source or two independent outlets say the event did not happen. False reports are left out of the event feed; they can appear in source history and claim matches.

Linking reports or gaining reposts never confirms a claim. Confidence is a 0–100 assessment, not a calibrated probability or a guarantee of truth. The v1 event response does not expose the review timestamp or administrator audit notes.

Source credibility over time

GET /v1/sources/{id} returns a stored reliability_score, confirmed and false claim counts, editorial notes, metric_history and recent_claim_outcomes. Reliability and settled-claim accuracy are different measures; reliability is not recalculated from the counts.

  • Accuracy: confirmed outcomes as a share of confirmed plus false, leaving out unresolved claims. null with no settled outcome. A small sample is limited evidence, not a validated track record.
  • Speed: how often the source published before every other independent source on the same events, and otherwise its median minutes behind the earliest one.
  • Originality: its share of reports that are not agency copies, and of reports no other independent source carried.
  • Corroboration: its share of reports at least an hour old that were corroborated or confirmed, with the shares settled false and still single-source. This is agreement, not proof.

Speed, originality and corroboration cover the last 90 days, carry their counts and sample, and are null below five reports.

Metric history records a snapshot when a source is created or its metrics change. recorded_at is when the snapshot was recorded, not when earlier reporting was assessed; no assessment is backdated. The endpoint returns up to 20 recent snapshots and a bounded set of recent claim outcomes, not every claim.

Independence & uncertainty

Distinct source IDs do not prove independent evidence: shared ownership, syndication, reposts and a common upstream report can connect separate publishers. Each cited source in an event response carries a group:

  • Copies of one agency story share the agency’s group.
  • Sources Statiq has recorded as one group or under one owner share it.
  • Any other source’s group is the website domain of its link, or of the source’s own address when the link has none.

verification.independent_outlets counts the distinct groups among the cited sources.

verification.confidence_calculation lists the groups counted when the confidence was set, so they can differ from the cited sources. There, storm reports from different kinds of observer (a trained spotter, law enforcement, the public) are separate groups, while every cited storm report shows the group weather.gov; a corroborated storm report can therefore show independent_outlets of 1. The calculation can also include evidence that is not cited publicly, and it leaves out sources added after it. Ownership or a common upstream report that Statiq has not recorded can still connect separate groups, so repeated coverage alone is not enough.

Locations use places the reporting names. Country-level placements can be approximate, and ambiguous or global reports can stay unmapped. Transit alerts are placed from the stops and routes they name, or at the agency’s city; county-wide weather alerts sit at the county centre and are labelled area-wide. GET /v1/events returns coordinates without their precision (GET /v1/local adds location_precision), so do not treat coordinates alone as exact incident locations.

Claim matching & corrections

  • POST /v1/verify_claim retrieves related reports and their recorded status; it does not verify a new claim.
  • No match means no close enough eligible report was found, not that the claim is false. An empty contradiction list does not mean no contradicting evidence exists.
  • Claim-status changes are recorded in an append-only history. GET /v1/conflict-events/{event_id} returns it for coded reports; elsewhere the API returns source snapshots and recent outcomes, not the full history or the administrator audit trail.
  • Keep citation URLs, event IDs and retrieval dates when you use evidence downstream, and re-check current records before consequential decisions.

Place data

Maps and town names come from open data: Natural Earth (public domain) and US Census places. Strategic places such as canals, straits, ports, border crossings and bridges come from Wikidata (CC0), with their extents from Natural Earth. A report placed on such a place sits at its centre and is labelled that way; it is not an exact incident location. The boroughs and districts of large cities that the place search offers, such as New York City's and London's, come from Wikidata (CC0) and US Census counties.

MCP server for AI agents

The MCP server gives an agent the same data as the REST API, with the same key. It speaks Streamable HTTP (MCP revisions 2024-11-05 to 2025-11-25) with plain JSON responses and no sessions, so any client that can send a header can connect. Every tool is read-only, and results carry status and confidence exactly as recorded. Every tool except geocode_place goes through the same scopes, quotas, plan delay and audit log as REST.

ToolReturns
search_eventsPublished events by region, category, time and bounding box, as GET /v1/events.
get_eventOne event by id, with its sources and corrections.
local_eventsEvents near a point, as GET /v1/local.
verify_claimReports matching a claim, as POST /v1/verify_claim.
get_sourceA source’s record, as GET /v1/sources/{id}.
get_storyA story and its member events, as GET /v1/stories/{id}.
conflict_events, conflict_event, conflict_trends, conflict_areas, conflict_movementsThe conflict endpoints below, as JSON (Professional and Complete).
geocode_placeCoordinates for a place name from Statiq’s place data. Uses no quota.

On Individual, MCP offers search_events, get_event, local_events, geocode_place only, with at most 20 calls a UTC day inside the monthly allowance and 10 results an answer. tools/list shows your plan’s tools; any other tool answers with a JSON error.

claude mcp add --transport http statiq https://statiq.sh/api/public/mcp \
  --header "Authorization: Bearer $STATIQ_API_KEY"
curl -s https://statiq.sh/api/public/mcp \
  -H "Authorization: Bearer $STATIQ_API_KEY" -H 'Content-Type: application/json' \
  -d '{"jsonrpc":"2.0","id":1,"method":"tools/list"}'

Connect an MCP client

Each client below connects with a Statiq API key. Keep the key in an environment variable, or in the client’s own secret prompt where it has one. Clients that cannot send a header, such as ChatGPT and claude.ai connectors without the Request headers option, can use the keyless sandbox below.

Claude Code

terminal

claude mcp add --transport http statiq https://statiq.sh/api/public/mcp \
  --header "Authorization: Bearer $STATIQ_API_KEY"

Cursor

~/.cursor/mcp.json (or .cursor/mcp.json in a project)

{
  "mcpServers": {
    "statiq": {
      "url": "https://statiq.sh/api/public/mcp",
      "headers": {
        "Authorization": "Bearer ${env:STATIQ_API_KEY}"
      }
    }
  }
}

VS Code (GitHub Copilot agent mode)

.vscode/mcp.json; VS Code asks for the key once and stores it

{
  "inputs": [
    {
      "type": "promptString",
      "id": "statiq-key",
      "description": "Statiq API key",
      "password": true
    }
  ],
  "servers": {
    "statiq": {
      "type": "http",
      "url": "https://statiq.sh/api/public/mcp",
      "headers": {
        "Authorization": "Bearer ${input:statiq-key}"
      }
    }
  }
}

OpenAI Codex CLI

~/.codex/config.toml

[mcp_servers.statiq]
url = "https://statiq.sh/api/public/mcp"
bearer_token_env_var = "STATIQ_API_KEY"

Gemini CLI

~/.gemini/settings.json. httpUrl is Streamable HTTP; url would mean SSE, which Statiq does not offer.

{
  "mcpServers": {
    "statiq": {
      "httpUrl": "https://statiq.sh/api/public/mcp",
      "headers": {
        "Authorization": "Bearer $STATIQ_API_KEY"
      }
    }
  }
}

Windsurf

Windsurf's mcp_config.json

{
  "mcpServers": {
    "statiq": {
      "serverUrl": "https://statiq.sh/api/public/mcp",
      "headers": {
        "Authorization": "Bearer ${env:STATIQ_API_KEY}"
      }
    }
  }
}

Zed

settings.json. Zed documents no variable expansion here, so the key is written in the file.

{
  "context_servers": {
    "statiq": {
      "url": "https://statiq.sh/api/public/mcp",
      "headers": {
        "Authorization": "Bearer stq_your_key"
      }
    }
  }
}

Claude Desktop

claude_desktop_config.json, through the mcp-remote bridge (Node.js). No space after the colon: some hosts split arguments at spaces.

{
  "mcpServers": {
    "statiq": {
      "command": "npx",
      "args": [
        "mcp-remote",
        "https://statiq.sh/api/public/mcp",
        "--header",
        "Authorization:${AUTH_HEADER}"
      ],
      "env": {
        "AUTH_HEADER": "Bearer stq_your_key"
      }
    }
  }
}

OpenAI Responses API

an entry of tools in a responses request. Headers are not stored by OpenAI, so send them with every request.

{
  "type": "mcp",
  "server_label": "statiq",
  "server_url": "https://statiq.sh/api/public/mcp",
  "headers": {
    "Authorization": "Bearer stq_your_key"
  },
  "require_approval": "never"
}

Claude Messages API (MCP connector)

the request body, with the header anthropic-beta: mcp-client-2025-11-20. The token is sent as Authorization: Bearer, which Statiq accepts.

{
  "mcp_servers": [
    {
      "type": "url",
      "url": "https://statiq.sh/api/public/mcp",
      "name": "statiq",
      "authorization_token": "stq_your_key"
    }
  ],
  "tools": [
    {
      "type": "mcp_toolset",
      "mcp_server_name": "statiq"
    }
  ]
}

MCP sandbox (no key)

Try the MCP tools without a key. https://statiq.sh/api/public/mcp/sandbox answers search_events, get_event, local_events and verify_claim, with the keyed server’s names and input schemas, from four recorded example reports. It never returns live events, and every answer is labelled recorded. It needs no key or account and calls no model. Each network may send 20 requests a minute and 200 a day. Every other tool needs a key on /api/public/mcp.

ChatGPT (developer mode)

a custom MCP app or connector. ChatGPT sends no custom headers and Statiq does not offer MCP OAuth, so ChatGPT connects to the sandbox only.

URL: https://statiq.sh/api/public/mcp/sandbox
Authentication: No authentication

claude.ai, Claude Desktop and mobile (custom connector)

Settings, Connectors, Add custom connector. Organisations with the Request headers option can add a keyed connector instead: URL /api/public/mcp, header authorization = Bearer stq_your_key.

URL: https://statiq.sh/api/public/mcp/sandbox
Authentication: none
curl -s https://statiq.sh/api/public/mcp/sandbox -H 'Content-Type: application/json' \
  -d '{"jsonrpc":"2.0","id":1,"method":"tools/call","params":{"name":"search_events","arguments":{}}}'

GET /v1/events

Lists published events, newest first. Scope: events:read.

ParameterMeaning
regionExact match on the event’s region, such as North America, Latin America, Europe or Global.
categoryOne of strike, troop_movement, protest, diplomatic, shipping, cyber, political, weather, outage, public_safety, transport, fire, health, civic, other.
sinceISO 8601 timestamp with offset.
min_confidence0–100.
bboxwest,south,east,north in degrees.
idsUp to 20 comma-separated event ids.
limit1–100, default 50.

Published means a recorded AI review or an audited confirmation. A single-source report can be published while its status is unverified. Reports settled false are left out. Each event carries:

  • id, title, summary, category, lat, lon, status, confidence and first_detected_at.
  • sources: each public source’s id, platform, reliability_score, group, url and posted_at.
  • verification: independent_outlets (distinct groups among the public sources); confidence_calculation (the method and recorded inputs behind the confidence, with reproduced; its groups are those counted when the confidence was set, so they can differ from the cited sources); basis (single_source, independent_outlets, settlement_squad or administrator_review); settled_by (squad or administrator once confirmed or false, otherwise null); verified_at; and a note (left out below).
  • corrections: each correction in force, with corrected_at, field and note.

Every response names its calculations version.

curl 'https://statiq.sh/api/public/v1/events?category=shipping&min_confidence=60&bbox=20,30,40,45' \
  -H "Authorization: Bearer $STATIQ_API_KEY"
{
  "events": [
    {
      "id": "00000000-0000-4000-8000-000000000001",
      "title": "Example: port operations disrupted",
      "summary": "An illustrative source-backed report.",
      "category": "shipping",
      "lat": 35,
      "lon": 30,
      "status": "corroborated",
      "confidence": 72,
      "first_detected_at": "2026-11-10T10:00:00Z",
      "verification": {
        "independent_outlets": 2,
        "independent_outlets_method": "statiq-independent-groups-v1",
        "confidence_calculation": {
          "method": "statiq-confidence-v1",
          "previous_confidence": 40,
          "groups": [
            {
              "group": "example.org",
              "reliability": 85
            },
            {
              "group": "example.net",
              "reliability": 75
            }
          ],
          "result": 72,
          "recorded_at": "2026-11-10T10:40:00Z",
          "reproduced": true
        },
        "basis": "independent_outlets",
        "settled_by": null,
        "verified_at": null
      },
      "sources": [
        {
          "id": "00000000-0000-4000-8000-000000000002",
          "platform": "news",
          "reliability_score": 85,
          "group": "example.org",
          "url": "https://example.org/report",
          "posted_at": "2026-11-10T09:58:00Z"
        },
        {
          "id": "00000000-0000-4000-8000-000000000003",
          "platform": "news",
          "reliability_score": 75,
          "group": "example.net",
          "url": "https://example.net/story",
          "posted_at": "2026-11-10T10:31:00Z"
        }
      ],
      "corrections": []
    }
  ],
  "calculations": "2026-11-12"
}

POST /v1/verify_claim

Finds up to five published reports that match a claim. Scope: claims:verify. Send {"text": "…"} with 10–2,000 characters of claim text; the body may be at most 10,000 bytes.

  • Matching is lexical and deterministic, with no AI call. It searches at most 500 recent eligible reports.
  • A report matches when at least three of the claim’s significant terms appear in its title or summary, and they cover at least half of the claim or make up at least 35% of the claim’s and the report’s terms together.
  • Each match carries the fields of GET /v1/events (shortened below) and retrieval_similarity: the share of the claim’s significant terms found, to three decimals.
  • Supporting and contradicting sources follow the matched report’s recorded status, not an inferred contradiction.
  • With no close enough report, result is no_match.

Responses are labelled "matching": "lexical". A match is retrieval, never a new verdict.

curl 'https://statiq.sh/api/public/v1/verify_claim' \
  -H "Authorization: Bearer $STATIQ_API_KEY" \
  -H 'Content-Type: application/json' \
  -d '{"text":"Port operations disrupted at Northport cargo terminal"}'
{
  "result": "matches",
  "matches": [
    {
      "id": "00000000-0000-4000-8000-000000000001",
      "title": "Example: port operations disrupted",
      "status": "corroborated",
      "confidence": 72,
      "retrieval_similarity": 0.5,
      "supporting_sources": [
        {
          "id": "00000000-0000-4000-8000-000000000002",
          "platform": "news",
          "reliability_score": 85,
          "group": "example.org",
          "url": "https://example.org/report",
          "posted_at": "2026-11-10T09:58:00Z"
        },
        {
          "id": "00000000-0000-4000-8000-000000000003",
          "platform": "news",
          "reliability_score": 75,
          "group": "example.net",
          "url": "https://example.net/story",
          "posted_at": "2026-11-10T10:31:00Z"
        }
      ],
      "contradicting_sources": [],
      "evidence_note": "Attached report status; retrieval never creates a verdict."
    }
  ],
  "matching": "lexical",
  "matching_method": "statiq-lexical-v1",
  "calculations": "2026-11-12",
  "limitations": "Lexical retrieval over recent eligible events (title and summary). Retrieval does not verify a claim."
}
{
  "result": "no_match",
  "matches": [],
  "matching": "lexical",
  "matching_method": "statiq-lexical-v1",
  "calculations": "2026-11-12",
  "limitations": "Lexical retrieval over recent eligible events (title and summary). Retrieval does not verify a claim."
}

GET /v1/local

Lists events near a point from the last 14 days, ranked by relevance. Scope: events:read.

ParameterMeaning
lat, lonThe point, in degrees. Required.
radius_miles1–50, default 25.
localityOptional. Adds county- and region-wide reports whose locality matches exactly, ignoring case (for example Cook County, IL).
limit1–100, default 25.

Relevance is confidence × max(0, 1 − distance ÷ radius) × 2^(−age in minutes ÷ 120), a two-hour half-life, returned with its relevance_inputs. Each event adds locality, location_precision, distance_miles and area_wide. Area-wide reports carry no distance. Ties, including every area-wide report, are newest first. An empty answer is a coverage gap, not an all-clear.

curl 'https://statiq.sh/api/public/v1/local?lat=41.88&lon=-87.63&radius_miles=10' \
  -H "Authorization: Bearer $STATIQ_API_KEY"

GET /v1/stories/{id}

Returns a public story and its member events (up to 100, in time order), each with its own public sources, plus history: the members’ recorded status changes. Scope: stories:read. Members follow your plan’s delay and history.

  • summary and citations are returned only while the members you can see are those the summary was written from (their titles, summaries, statuses and confidence) and every citation is one of them. Otherwise summary is null.
  • summary_held: true marks a quiet single-source story whose summary can be up to 12 hours older than its newest members; see summary_generated_at.
  • This endpoint never calls a model.

GET /v1/sources/{id}

Returns a public source’s record by its UUID. Scope: sources:read. A source that is not public answers 404. What each measure means is under Source credibility over time.

  • reliability_score and editorial_notes: the stored assessment.
  • claims_confirmed, claims_false and outcomes: settled outcomes, and how many in the recent sample the settlement squad or an administrator settled.
  • components: accuracy from the settled outcomes, then speed (first_share, median_minutes_behind and each report’s minutes_behind), originality (own_reporting_share, exclusive_share) and corroboration (corroborated_share, false_share, single_source_share) over the last 90 days. Each carries its method, counts and sample; shares are null below five reports.
  • assessment: the latest source assessment, or null when the source is not yet assessed: measured metrics, and model-judged scores kept only with quotes from the sampled headlines. Neither changes reliability.
  • metric_history (up to 20 snapshots) and recent_claim_outcomes.
curl 'https://statiq.sh/api/public/v1/sources/00000000-0000-4000-8000-000000000002' \
  -H "Authorization: Bearer $STATIQ_API_KEY"
{
  "id": "00000000-0000-4000-8000-000000000002",
  "display_name": "Example source",
  "platform": "news",
  "reliability_score": 85,
  "claims_confirmed": 3,
  "claims_false": 0,
  "editorial_notes": null,
  "assessment": null,
  "outcomes": {
    "confirmed": 3,
    "false": 0,
    "settled_by_squad_in_sample": 3,
    "settled_by_administrator_in_sample": 0
  },
  "components": {
    "accuracy": 100,
    "accuracy_method": "statiq-source-accuracy-v1",
    "speed": {
      "method": "statiq-source-speed-v1",
      "sample": 6,
      "first": 2,
      "first_share": 33,
      "minutes_behind": [
        0,
        0,
        4,
        12,
        31,
        58
      ],
      "median_minutes_behind": 8
    },
    "originality": {
      "method": "statiq-source-originality-v1",
      "sample": 20,
      "own": 17,
      "exclusive": 6,
      "own_reporting_share": 85,
      "exclusive_share": 30
    },
    "corroboration": {
      "method": "statiq-source-corroboration-v1",
      "sample": 18,
      "corroborated": 9,
      "false": 0,
      "single_source": 7,
      "corroborated_share": 50,
      "false_share": 0,
      "single_source_share": 39
    }
  },
  "metric_history": [],
  "recent_claim_outcomes": []
}

GET /v1/conflict-events

Lists conflict and protest events coded by Statiq from its own reports, newest event date first: event and sub-event type, actors, place, date and fatalities as reported (null when not stated). Scope: events:read. Professional and Complete only: an Individual key is refused before anything is counted. Each call counts once against the monthly allowance.

ParameterMeaning
days1–365, default 7: event dates from today minus that many days.
since, untilYYYY-MM-DD, at most 366 days a call. An earlier since moves to your history start, and the answer says so.
countryFull English name, exact match.
regionStatiq’s region name.
typeBattles, Explosions/Remote violence, Violence against civilians, Protests, Riots or Strategic developments.
statusunverified, corroborated, confirmed or false: the report’s current status.
locatedtrue or false.
cursorThe previous page’s next_cursor.
limit1–500, default 100.
formatjson (default) or csv.

Each row carries the code and:

  • verification: as in GET /v1/events, with the recorded statiq-confidence-v1 or statiq-settled-v1 calculation and reproduced.
  • evidence_groups: the independent groups of the public receipts; their number is independent_outlets (statiq-independent-groups-v1).
  • corroboration (statiq-corroboration-window-v1), as of the answer: confirmed, corroborated, official, checking (a single-source report’s first 15 minutes, with minutes_left), single_source, or false for a report settled false.
  • code: method (the coder version), model, coded_at, status_at_coding and quotes_checked, the fields stored with a quote found verbatim in the receipts. conflict-coder-v4 stores a fatality figure or an actor only with such a quote; older codes carry current_method false.
  • Codebook fields, null on codes made before conflict-coder-v4: inter1 and inter2 (each actor’s type, from 1 state forces to 8 external or other forces, by its goals and structure); interaction (the two types, lower first; a one-sided event is its type followed by 0) with interaction_name; assoc_actor_1 and assoc_actor_2 (each with a checked quote); civilian_targeting (only when civilians were the main or only target); time_precision (1 a day, 2 a period or no day given, 3 a month); and tags (“crowd size=…” for Protests and Riots, “no report” when no receipt states a size). disorder_type follows from the sub-type on every code.
  • admin1 and admin2 come from Statiq’s place data, never the model: in the US, the Census state and county; elsewhere a region only when a receipt also names it.

The answer also carries total (every code matching the filters), next_cursor (keyset by event date, then id), since, until, history_from, calculations and methods. Protest and riot codes, and any code on a protest report, are withheld when the report carries a post from a platform whose terms forbid protest monitoring.

format=csv returns the same page as CSV, with the columns event_date, year, time_precision, disorder_type, event_type, sub_event_type, actor1, assoc_actor_1, inter1, actor2, assoc_actor_2, inter2, interaction, civilian_targeting, iso, region, country, admin1, admin2, location, latitude, longitude, geo_precision, source_scale, notes, fatalities and tags, then statiq_event_id, status, confidence, evidence_groups, first_detected_at, coded_at and method. The first line, starting with #, names the data and its method. fatalities is empty when no receipt states a figure, never 0 by default; region uses Statiq’s region list; evidence_groups counts the independent sources behind the code. The total, the next cursor and the label come in the X-Total-Count, X-Next-Cursor and X-Statiq-Label headers.

The console’s Conflict page shows the same coding:

  • Filters by range (7, 30 or 90 days, or custom dates), region, country and event type. Filters and sort stay in the page address, so a view can be shared; CSV and JSON exports apply them.
  • Trends (events by type and region, reported fatalities as stated, and changes by region and country) are counted as the signed-in reader.
  • A change compares the range’s length in complete days, ending yesterday at the latest, with the same number of days just before. It appears only when coding covered both; otherwise the page says when coding began.
  • A code dated after its report was first detected (later than that date anywhere, at UTC+14) or more than 30 days before it stays in the table and exports but is left out of the trends, and the page says how many.
  • The map steps through the range in 1-, 3- or 7-day windows and draws located reports only. It shows where Statiq’s coded reports were made, not front lines, control or safety.
curl 'https://statiq.sh/api/public/v1/conflict-events?days=7&country=Sudan&status=corroborated' \
  -H "Authorization: Bearer $STATIQ_API_KEY"

GET /v1/conflict-events/{event_id}

Returns one coded report by its id, with its full verification trail. Scope: events:read; Professional and Complete only; one call each.

  • Its public receipts as links: source, reliability, group, link and time. No receipt text is returned.
  • The claim history in order, which is append-only.
  • Every corroboration and settlement action with the inputs it was calculated from, reproduced (the published formula gives the same result) and undone_at when undone.
  • settled_by (the settlement squad or an administrator), official_source, life_safety and the corroboration window.

404 when the report is not coded or your key cannot see it: plan delay, history, the publication rule, or a source platform’s terms.

curl 'https://statiq.sh/api/public/v1/conflict-events/00000000-0000-4000-8000-000000000001' \
  -H "Authorization: Bearer $STATIQ_API_KEY"

Returns two series of conflict counts, never mixed. Scope: events:read; Professional and Complete only; one call each. Query: days (1–365, default 30, today included) or since and until; region, country and type.

  • statiq: Statiq’s own coded history (conflict-trends-v1), counted exactly as the Conflict page counts it for a paid reader. Daily events by type and region; reported fatalities per day with how many codes gave a figure (as stated, never estimated); countries with the like-for-like change and its coverage note; and left_out, the codes dated after detection or more than 30 days before it, kept out of every trend. total equals counted plus left_out.
  • gdelt: “GDELT Event Database, not Statiq reports”. Monthly counts of GDELT’s notable events (root events with a place, first seen within three days, reported by 3 or more sources or in 5 or more articles) by CAMEO root: 14 Protest, 15 Exhibit force posture, 17 Coerce, 18 Assault, 19 Fight, 20 Use unconventional mass violence. For the country you name, or for the world with the 25 busiest countries. Machine-coded by GDELT from news, never reviewed by Statiq and never added to Statiq’s counts; region and type do not apply. Counts only, never GDELT’s rows. Every answer carries the citation and link GDELT’s terms ask for (https://www.gdeltproject.org/).
curl 'https://statiq.sh/api/public/v1/conflict/trends?days=90&country=Sudan' \
  -H "Authorization: Bearer $STATIQ_API_KEY"

GET /v1/conflict/areas

Returns the active conflict areas: up to 12 countries, each with counts, a box and a cited recap. Scope: events:read; Professional and Complete only; one call each. No query parameters.

  • Active (conflict-active-v1): at least 10 violent coded events (Battles, Explosions/Remote violence, Violence against civilians) dated in the last 7 UTC days, with dates plausible for their reports; the 12 busiest. Only published, non-duplicate reports not settled false, past your plan’s delay and inside its history, are counted.
  • Counts: violent, total, unverified, corroborated, confirmed, located, precise and by type.
  • Box (conflict-box-v1): the 10th to 90th percentile of the precisely located codes, padded by the larger of half a degree and a tenth of the span, with box_inputs to recompute it. Null below three precise codes.
  • Recap (conflict-recap-v1): written by an AI model from Statiq’s codes, titles and statuses, never from receipts. Each sentence lists the report ids it rests on. Corroborated and confirmed reports come first; a single-source report is cited only when the area has fewer than five of those, and is always called single-source. Fatalities appear only as one report gave them, never added up. A recap never forecasts or describes front lines, control or intent.
  • A sentence citing a report your key cannot see is left out and counted in sentences_withheld. A recap is rewritten after its area’s codes change; until then the previous one is returned with stale: true and codes_changed_since.
curl 'https://statiq.sh/api/public/v1/conflict/areas' \
  -H "Authorization: Bearer $STATIQ_API_KEY"

GET /v1/conflict/movements

Returns land, sea and air activity for one active area, as counts. Scope: events:read; Professional and Complete only; one call each.

ParameterMeaning
areaAn active area’s country as /conflict/areas names it, ignoring case. Any other value answers 404 with the active areas.
daysDays of land moves: 1–30, default 7.
hoursHours of hourly counts: 1–168, default 24.
  • land (conflict-land-moves-v1): land moves reported in coded reports of territory overtaken by a non-state actor or handed over without violence. Each has its report id, title, status, date and place as written. Their placement is not accurate enough to map, so moves never carry coordinates or a direction.
  • sea (vessel-counts-v1): cargo ships (AIS types 70–79) and tankers (80–89) reporting now, and how many are waiting (at anchor, or under 0.5 kn and not moored). Counted over the area’s box when it is small enough to listen to, otherwise only in the AIS-covered ports and straits inside it; coverage says which.
  • air (flight-classes-v1): aircraft broadcasting now, by class: emergency, tanker, surveillance, airlift, military, firefighting and cargo (holding or circling). Everyday airliners, private and medical flights, VIP and head-of-state transport, and aircraft flagged not to be displayed are never counted.
  • history (conflict-area-counts-v1): hourly counts by class. A null value was not counted that hour.

Counts only: no ship or aircraft identifier, name, position or trail, and never intent. An empty count is not an all-clear: most military aircraft and naval vessels do not broadcast. Ship positions: AISStream. Aircraft positions: © adsb.lol (ODbL).

curl 'https://statiq.sh/api/public/v1/conflict/movements?area=Yemen' \
  -H "Authorization: Bearer $STATIQ_API_KEY"

Example requests and responses on this page are illustrative, not live data.