150+
autonomous collectors
sweeping the signal families below on a schedule, normalized to one record contract — a conflict event and a sanctions designation carry provenance the same way
Advanced Research Division
the-synactic
One instrument. A world of signals. Every claim accounted for.
the-synactic is Praxen's self-hosted geopolitical-intelligence instrument: a realtime open-source-intelligence baseline, collected at machine scale and exposed as typed, machine-readable intelligence that an AI agent can read directly. It is built for defensive analysis — situational awareness, risk screening, and research — and it is read-only by charter.
Rent your intelligence, or own it
Rented intelligence looks like this: a dashboard you do not control, metered quotas on your own questions, sources that change without notice, and figures you are asked to trust because the interface looks confident.
the-synactic is the alternative. The collection fleet runs on hardware in the building. The data plane lives in a cache no third party custodies. And every figure that reaches a deliverable carries its provenance — because the difference between intelligence and content is being able to say where a number came from.
The posture
60+
typed MCP tools
the agent-facing surface — each self-describing, projectable server-side, and callable without reading a page of documentation first
Read-only by charter
no order, no trade, no transfer — the boundary is structural, not a setting someone can flip
Self-hosted end to end
collection, data plane, and tool surface all run on owned hardware; no third-party SaaS on the critical path
Every record carries its source and its clocks — provenance, fetched_at, and an explicit staleness state.
The baseline
Continuous collection across the signal families an analyst actually has to hold at once. Collected on a schedule, normalized to one record contract, and kept queryable.
Conflict & security
Conflict events, military posture, and hotspot escalation, scored and time-stamped rather than narrated.
Sanctions & compliance
Sanctions programs and designated-entity exposure, resolvable per country and per entity type.
Movement & advisories
Travel and health advisories, displacement signals, and population exposure, each carrying the source that issued it.
Trade & chokepoints
Supply-chain status and maritime chokepoint throughput against published reference baselines.
Energy & macro
Energy flows and macroeconomic indicators, aligned to the same freshness contract as everything else.
Cyber & natural hazard
Cyber threat signals, natural-disaster activity, and climate anomalies, collected continuously.
Built to be read by machines
The consumption model is MCP-first. The instrument does not publish a dashboard and hope an agent can scrape it; it publishes typed tools an agent can discover, describe, and call.
Self-describing schemas
Every tool carries its own input and output contract, so an agent can reason about a call before making it.
Server-side projection
JMESPath projection runs before the response leaves the server, so a context window carries the fields the task needs and nothing else.
Compressed descriptions
Tool listings stay small by default; the full definition is one deliberate call away rather than always resident.
Quota-exempt discovery
Exploring the tool surface does not consume the budget for using it. Agents can orient before they spend.
What an agent sees
A country-risk read, as the calling agent receives it. The field shape is the real contract; the values below are synthetic and the country is a placeholder.
Illustrative response synthetic values
{
"countryCode": "XX",
"countryName": "Example State",
"cii": {
"combinedScore": 61,
"staticBaseline": 44,
"dynamicScore": 17,
"trend": "TREND_DIRECTION_RISING",
"components": {
"newsActivity": 6,
"militaryActivity": 22,
"geoConvergence": 3,
"ciiContribution": 0
},
"methodologyVersion": "v8",
"computedAt": 1785600000000,
"advisoryLevel": "reconsider-travel",
"advisoryProvenance": "fallback"
},
"sanctionsActive": true,
"sanctionsCount": 112,
"fetchedAt": 1785600000000,
"upstreamUnavailable": false
}- advisoryProvenance
- The advisory came from a fallback source, not the primary issuer. The agent is told, so the analyst can discount it.
- upstreamUnavailable
- An explicit unavailability flag. A degraded input announces itself instead of silently resolving to a plausible-looking number.
- fetchedAt · computedAt
- Two clocks, not one: when the input was retrieved, and when the score over it was computed. Staleness is measurable rather than assumed.
- methodologyVersion
- The scoring contract is versioned, so a figure in a report six months old can be re-derived under the method that produced it.
Nothing on this page is a live feed. The block above is rendered as static text to show record structure — no request is made when you load it.
The engineering
Collection is the commodity. The discipline around the record is the instrument.
Provenance on every record
Every value carries where it came from and when it was fetched. Where a primary source was unreachable and a secondary stood in, the record says so in a field — not in a footnote nobody reads. A number without a source is not intelligence; it is content.
Honesty patterns
Unavailable is not the same as failed, and neither is the same as zero. The instrument distinguishes all three and discloses which one it is. An honest empty answer beats a confident fabrication, and the schema is built so that empty can be said out loud.
Contract truthing
Declared schemas are verified against captured real payloads, not against themselves. The regression suite asserts presence and absence — a validator that cannot say no proves nothing. Schema drift is caught at build time instead of in a client deliverable.
Sovereignty
Fully self-hosted under Docker. No third-party SaaS sits on the critical path, so no vendor's outage, quota change, or deprecation notice decides whether the baseline is available.
What it will not do
The instrument observes. It will never place an order, execute a trade, or move money — not as a configuration choice, but as a charter. Execution authority belongs to downstream systems operating under their own authorization, and the boundary is deliberate: an instrument that can act is an instrument whose readings you have to second-guess.
It is a capability Praxen operates, not a data product or a resale service.
Lineage
Built on the excellent WorldMonitor project by Elie Habib — an open-source, real-time global intelligence dashboard released under AGPL-3.0, with a hosted instance at worldmonitor.app. Both the project and that hosted service are his, not Praxen's. Praxen's layer is a private engineering fork; upstream components remain under their original licenses.
Own your instrument
Praxen builds the machines that build the briefings. If your organization is making decisions on rented dashboards and cannot say where the numbers came from, that is an engineering problem, and it is the kind Praxen Development takes.
The problem, the constraint, and what shipping looks like for you: chris@cld-dev.io