fleet-geo
Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type
@followed by the MCP server name and your instructions, e.g., "@fleet-geo where does gpt-4o actually run?"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
fleet-geo — an MCP that answers "where does this model actually run?"
A zero-dependency MCP server (stdio, Node >= 20) that resolves the physical location of inference for each model in the Orca fleet, and refuses to guess.
Written because a prompt rots: providers change, endpoints move, new lanes appear. A tool that re-derives the answer from live data does not.
fleet-geo-mcp/
server.mjs MCP stdio transport + 10 tools
src/observe.mjs observe_endpoints: proxy, OS socket table, strace SNI/DNS (no decryption)
src/doh.mjs DNS over HTTPS (ISP-independent evidence)
src/sources.mjs Cymru ASN, RDAP, AWS/GCP prefix tables, PeeringDB, crt.sh, globalping, headers
src/geo.mjs CIDR math, distance, RTT floor, region anchors, hostname fingerprints
src/classify.mjs evidence -> four planes -> confidence + contradictions
src/catalog.mjs the only hand-maintained data, plus record_fact
catalog.json provider seeds, all marked hypothesis until cited
fleet.json the lane table (W1-W7, C, B, T) - adding a lane is a one-line edit
config.json entitled-host allowlist + limits
test/selftest.mjs 41 checks (offline strict, online tolerant)
test/observe.test.mjs 11 observe_endpoints checks, offline on every OS
test/stdio-check.mjs protocol surface, replayed from a recorded session
dsh-bundle/ the DSH wrapper, shipped with the serverRuns on every OS
Windows, macOS and Linux, on Node.js 20 or newer, with zero dependencies. Every tool
uses only Node built-ins and public HTTPS/DNS sources, so it behaves the same on all three.
The only OS-specific part is observe_endpoints, which picks the native tool for each OS
and never needs root or admin rights:
Windows | macOS | Linux | |
all lookup tools ( | yes | yes | yes |
| yes | yes | yes |
|
|
|
|
| via the proxy | via the proxy | proxy + |
CLI shims ( | run through the shell | direct | direct |
process cleanup on timeout |
| process group | process group |
Per-OS differences in how a lane is launched live in fleet.json as
observe.argvByPlatform (for example agy needs a pseudo-TTY via script on Linux and
macOS). Shell examples below are PowerShell; the same commands work unchanged in bash or zsh.
Related MCP server: Proof Mode
The one idea
Four planes, never merged.
Plane | What it is | Why it matters |
CONTROL | the base URL the CLI actually calls | may be a router, not the model author |
EDGE | where TLS terminates | nearest to you; almost never the model |
ORIGIN | where weights execute | this is the answer |
LEGAL | entity + residency commitment | what a DPO accepts; constrains origin |
Every naive geolocation answer collapses these into one and reports the Cloudflare PoP down the street as "the datacenter".
Tools
Tool | Use |
| one lane: model / lane name / base_url -> full location card |
| every lane in one sweep -> table + roll-up |
| one hostname -> CNAME hops, ASN, cloud region, RDAP |
| one IP -> ASN, published cloud region, PeeringDB facilities |
| crt.sh subdomains, regional ones surfaced |
| distributed probes + the RTT floor falsifier |
| headers from an entitled endpoint -> edge PoP only |
| cite a provider page with a quote -> tier-A anchor |
| run a lane's one-line probe and record the hosts and IPs it really connects to |
| precedence, planes, confidence, falsifier |
observe_endpoints
Config files say what a CLI should call; this records what it does call. Each lane in
fleet.json carries an observe.argv (with optional argvByPlatform and a git fixture
for jev). With run:true the tool runs that one prompt and watches it through three layers:
Layer | Linux | macOS | Windows | Sees |
logging CONNECT proxy ( | yes | yes | yes | hostname, port, bytes per tunnel |
OS socket table for the process tree |
|
|
| every remote IP, proxy or not |
| yes | - | - | TLS SNI hostname, DNS questions, every connect |
Nothing is decrypted: the hostname comes from the CONNECT line, the cleartext SNI or the
DNS question. Every destination IP then gets lookup_prefix-style facts and an EDGE /
ENDPOINT label. The result is CONTROL-plane evidence, saved under observations/.
It ends at the TLS edge and never proves the origin. Without run:true it only prints
the commands, because each lane costs one model call.
The answer shape
{
"planes": { "control": {...}, "edge": {...}, "origin": {...}, "legal": {...} },
"location": { "country": "KR", "region": "ap-northeast-2", "facility": null,
"precision": "region" }, // facility > metro > region > country > unknown
"confidence": "ANCHORED", // ANCHORED | CORROBORATED | INDICATED | OPEN
"confidenceWhy": "1 tier-A anchor(s): cloud-published-region",
"contradictions": [], // e.g. an RTT below the fiber floor
"evidence": [ { "id": "E2", "tier": "A", "source": "aws ip-ranges.json",
"observed": "3.5.140.1 falls in 3.5.140.0/22, published as AWS region ap-northeast-2" } ],
"nextProbes": ["Region is anchored. A facility needs documentary evidence: ..."]
}precision is the honest word for "exact". A tool that returns a street address for a
Cloudflare-fronted API is lying; one that returns region: ap-northeast-2, precision: region, confidence: ANCHORED is not.
Precedence (highest first)
provider documentation (cited via
record_fact) / region encoded in the hostnamethe cloud's own published prefix table (AWS
ip-ranges.json, gstaticcloud.json)PeeringDB facility for the serving ASN
RDAP / ASN registration
commercial geoIP databases
latency inference
Rule 2 is the sharpest trick: when an address falls inside a published AWS or GCP prefix, the cloud itself tells you which region owns it. That beats every geoIP product, and it costs nothing.
The falsifier
Fiber runs at ~200,000 km/s, so a round trip costs ~1 ms per 100 km of path. Real paths are longer than the great circle, so measured RTT normally sits above the floor. A measurement far below it is physically impossible — and therefore a hard test:
// probe_latency from Helsinki @1.4 ms, claimed origin Ashburn (7000 km)
{ "verdict": "FALSIFIED",
"rows": [{ "impliedMaxDistanceKm": 140, "distanceToClaimedOriginKm": 7000,
"floorMs": 70, "violation": { "floorMs": 70, "hardFloorMs": 53 } }] }It can only ever disprove a claim. That is exactly what a confident answer needs.
Guardrails in code, not just in prose
Entitlement gate.
inspect_edgerefuses any host absent fromconfig.jsonentitledHosts. Passive lookups stay open; traffic does not.Pacing. Per-source minimum intervals. These are free shared services.
Secret hygiene. Auth headers are stripped before they can be echoed.
Provenance gate.
record_factrejects a fact without anhttp(s)URL and a >=12 char verbatim quote. An agent cannot promote its own guess to a fact.No invasive technique. DNS, RDAP, published tables, PeeringDB, crt.sh, ICMP from a third-party probe network. No port scanning, no whois TCP/43, no auth walls.
Dynamic by construction
Nothing about a provider is hardcoded except what only a human can supply: who publishes the model, and what their documentation claims.
New endpoint?
trace_endpointresolves it now.New provider?
record_factwithpublisher+match[]adds it.Provider moves? Live DNS and published ranges follow; the card changes.
Documentation ages? Every fact carries
verified_at;staleness()flags anything paststaleAfterDays(90) and prints "never verified" for the rest.
Honest limits
For CDN-fronted APIs the origin is not externally observable — the card returns
OPENplus the exact probe that would settle it.Azure publishes no queryable service-tag JSON any more, so Azure region comes from the resource itself (
az cognitiveservices account show ... --location); the tool says so rather than guessing.Passthrough aggregators (OpenRouter, Together, Fireworks) return a
providerfield per response. That is authoritative for the serving entity; the tool names it as the next probe instead of inventing it.IPv6 addresses skip the ASN lookup in v1 (IPv4 only);
observe_endpointslabels themUNKNOWNrather than presuming they are unicast endpoints.observe_endpointssees where a CLI connects, which ends at the TLS edge. A CLI that hands traffic to a local gateway outside its process tree is reported asLOCAL_GATEWAYwith the loopback ports, not as a location.
Develop
cd C:\DEV\fleet-geo-mcp
node server.mjs --selftest # 41 checks; offline strict, network tolerant
node test/observe.test.mjs # 11 observe_endpoints checks (or: npm test for both)
node server.mjs --tool observe_endpoints --args '{"lane":"all"}' # plan only
node server.mjs --tool observe_endpoints --args '{"lane":"W7","run":true}' # one model call
node server.mjs --list-tools
node server.mjs --tool fleet_status --args '{"limit":3}'
node server.mjs --tool locate_model --args '{"lane":"W7"}'
# full stdio session (shell redirection, not a piped child process)
cmd /c "node server.mjs < test\fixtures\stdio-requests.jsonl > test\fixtures\stdio-responses.jsonl"
node test/stdio-check.mjsCompanion
docs/PROMPT.md is the analyst prompt this server replaced: a static spec
that rots the moment a provider changes. Kept as the design rationale - the four planes, the
evidence ladder, the falsification pass - for when the tool's output needs arguing in prose.
Install
Two routes - a DSH bundle or a single patch row - both in INSTALL.md.
Placeholders there map to $env:DSH_NODE_EXECUTABLE, %USERPROFILE%\.dsh\profiles\<profile>,
and this checkout.
License
MIT - see LICENSE.
Available Tools
10 toolsexplain_rulesExplain the rulesA
Return the source precedence, plane definitions, confidence states and the RTT falsifier used by every other tool.
Inputs: no named parameters.
find_regional_endpointsFind regional endpointsA
Search certificate transparency logs for subdomains of a domain and surface ones that look regional (-eu, -jp, -sg, -cn ...). Providers often run regional endpoints that no documentation mentions.
Inputs: domain (required), limit.
fleet_statusWhole-fleet location sweepA
Run locate_model across every catalog provider that has a known API host and return one summary table: publisher, serving entity, edge operator, country/region, precision and confidence per lane. The one call to re-run whenever the fleet or a provider changes.
Inputs: deep, lanes, limit.
inspect_edgeInspect edge headersA
Read response headers from an ENTITLED endpoint only (config.json allowlist) and report the TLS-terminating PoP. Explicitly an edge-plane fact: it never establishes where inference ran.
Inputs: url (required).
locate_modelLocate a modelA
Resolve where a model's inference physically runs. Accepts a model id, a fleet lane/CLI name, or an explicit base_url. Returns a four-plane location card (control/edge/origin/legal), the serving chain, evidence rows with sources, a confidence state, contradictions, and the next probe that would tighten the answer. Re-derives everything from live DNS, ASN, cloud prefix tables and PeeringDB, so it tracks provider changes without redeploying.
Inputs: lane, model, base_url.
lookup_prefixLook up an IPB
Identify a single address: ASN and org, the cloud's own published region for the prefix, RDAP registration, and PeeringDB facilities. The published cloud region outranks any geoIP database.
Inputs: ip (required), include_facilities.
observe_endpointsObserve which hosts a lane really callsA
Run a lane's one-line probe prompt (fleet.json observe.argv) and record every outbound connection: a logging CONNECT proxy (every OS), the OS socket table for the process tree (ss / lsof / netstat), and on Linux strace for TLS SNI and DNS names. Nothing is decrypted. Settles the CONTROL plane only - observed hosts end at the TLS edge, never the inference origin. Spends one model call per lane, so it only runs with run:true; without it, it returns the commands it would run.
Inputs: lane (required), run, locate, timeout_s.
probe_latencyProbe latency and test the RTT floorB
Measure a host from distributed vantage points and test the physical RTT floor. Returns per-probe implied maximum distance, an anycast verdict, and a FALSIFIED verdict when a measurement is impossible for the claimed origin.
Inputs: host (required), limit, expected_region, expected_country, +1 more.
record_factRecord a documented factA
Add a provider residency or facility fact to the catalog, citing a source URL and a verbatim quote. This is how the catalog learns: a verified fact becomes a tier-A anchor and lifts the matching card to ANCHORED. Pass publisher + match[] to add a provider that is not in the catalog yet.
Inputs: kind (required), quote (required), provider (required), source_url (required), +8 more.
trace_endpointTrace an endpointA
Resolve one hostname into a serving chain: CNAME hops, addresses, ASN and org, registry RDAP, and the published cloud region of each address. Flags CDN/anycast addresses so an edge PoP is never mistaken for an origin.
Inputs: host (required), max_ips.
Tool Schema Changelog
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
TDQS
Scored across 10 tools
Most tools target clearly different objects: locate_model (model→location card), trace_endpoint (hostname→serving chain), lookup_prefix (IP→ASN/region), fleet_status (batch roll-up). The closest pairs are trace_endpoint vs lookup_prefix (hostname vs address, both emitting ASN/RDAP/cloud region) and inspect_edge vs observe_endpoints (both probes, different planes), but the descriptions explicitly disambiguate by input type and plane.
All names are snake_case and almost all follow verb_noun (inspect_edge, record_fact, probe_latency, locate_model, trace_endpoint, lookup_prefix, explain_rules). The one deviation is fleet_status, a noun-only name, which breaks the pattern slightly but remains readable in context.
Ten tools is well within the sweet spot and each maps to a distinct investigation step (edge, control, origin, legal, catalog write, discovery, falsification, meta). No tool feels redundant filler, and none of the planes is served by more than one or two tools.
Coverage is strong across the four planes: inspect_edge (edge), observe_endpoints (control), trace_endpoint/lookup_prefix/locate_model (origin), record_fact (catalog write), plus fleet_status and explain_rules. The notable gap is the catalog read/delete side — record_fact writes facts but there is no tool to list, inspect, or retract a fact, so bad entries cannot be corrected without re-recording.
Maintenance
Related MCP Connectors
MCP server for building and testing AI agents with multi-model experimentation and insights.
Experimental MCP server for current empirical verification of explicit public HTTPS endpoint claims.
MCP server providing access to the Scorecard API to evaluate and optimize LLM systems.
MCP server for OnceAsk, the AI-native current-address layer for people and agents.
Related MCP Servers
- AlicenseBqualityDmaintenanceMCP server for achieving peak inference performance by detecting drift between code and runtime behavior, comparing against benchmarks, and providing optimization templates.8Apache 2.0
- FlicenseNot gradedqualityCmaintenanceAn MCP server implementing a multi-phase backend for structured model interactions, bounded compilation, exact-hash confirmation, and Codex-run provenance, with trust-separation hardening.1-
- FlicenseAqualityDmaintenanceA self-hosted MCP server offering provider-independent AI routing, deterministic verification, and cryptographically signed, tamper-evident audit trails for every response.81-
- AlicenseNot gradedqualityBmaintenanceThis MCP server provides a stateful, resettable, verifiable API runtime that gates every tool call, enabling agents to run long workflows against provider-shaped environments without live provider write access. It records decisions, side effects, and outcome evidence for replayable, verifiable benchmark runs.Apache 2.0