Geospatial AI Infrastructure

The agent infrastructure for the physical world.

The only orchestration framework where every agent action is anchored to a real location, a real satellite observation, and a cryptographic proof.

H3
Cell-level precision
Ed25519
Every observation signed
< 5 min
Satellite to verified map
0.18%
Cloud cover — live Sentinel-2 tile
AGENT_01 851f9ebfffffff AGENT_02 confidence: 0.94 AGENT_03 h3_res: 7 · flood ✓ PROOF OF OBSERVATION H3:8928308280fffff · Ed25519 signed
The problem

Existing orchestrators are blind to the physical world.

Today's orchestration frameworks are built for digital space. They call APIs, process text, generate outputs — but have no concept of where an agent physically is, what it physically observed, or how to prove any claim about the real world. GEIANT fills that gap.

The industry baseline
Digital space only

LLM agents that call APIs, process text, generate outputs. No spatial context. No physical-world anchoring. No way to prove a claim about the real world.

  • No concept of geospatial location or jurisdiction
  • No satellite or sensor data integration
  • Claims about the physical world are unverifiable
  • Cannot operate in regulated spatial contexts
  • No proof of observation — any agent can assert anything
GEIANT
Grounded in the physical world

Every agent action is anchored to an H3 cell, a signed satellite observation, and a cryptographic proof chain. Jurisdiction-aware. Verifiable. Auditable.

  • H3-quantized spatial context for every agent
  • Live Sentinel-2 satellite tile integration (Planetary Computer)
  • Prithvi-EO-2.0 flood and event classification
  • Ed25519-signed perception chains — every claim is provable
  • Delegation model: each agent has jurisdiction over specific cells
  • Field operations anchored to physical reality — NRRP-grade audit trails
The perception chain

Five steps from Earth to proof.

Every GEIANT agent action follows the same verifiable pipeline. The full chain is stored in Spatial Memory — cryptographically signed, queryable, auditable.

H3
Spatial context

Agent receives an H3 cell index at resolution 5–8. This defines its operational territory. Delegation chains enforce jurisdiction.

SAT
Satellite query

GEIANT queries Microsoft Planetary Computer for the latest Sentinel-2C tile over the cell. Filters for cloud cover, acquisition date, band availability.

INFER
Perception inference

Prithvi-EO-2.0 classifies the tile — flood detection, land cover, anomaly scoring. Clay embeddings encode semantic meaning. WxC adds weather context.

MEM
Spatial Memory write

The full perception chain — tile ID, model ID, cell, timestamp, confidence, class counts — is written to PostGIS with version history.

SIGN
Agent signature

The agent signs the perception result with its Ed25519 key. This is the proof of observation — cryptographically tied to the agent's GNS identity.

What gets built on GEIANT

Four problems only physical-world agents can solve.

Real deployments. Real infrastructure. Real deadlines.

M+
● Live
Grid infrastructure
Smart meter monitoring at national scale

A major grid operator deploys GEIANT agents across millions of smart meters. Every anomaly alert is backed by a satellite flood/event classification and weather context — no false positives without physical proof.

Outcome: Zero unverified alerts. Every escalation includes satellite tile ID, acquisition timestamp, flood pixel percentage, and agent signature. Full audit trail for regulatory compliance.
GPS
Logistics & compliance
Jurisdiction-aware fleet operations

Autonomous agents track vehicles across regulatory boundaries. Proof-of-Trajectory from the GNS Protocol confirms physical presence at each waypoint — GPS spoofing is cryptographically detectable.

Outcome: Cross-border compliance automated. The agent knows which ruleset applies based on verified H3 cell — not just GPS coordinates that can be fabricated.
<5m
Disaster response
Satellite-to-verified flood map in minutes

After a flood event, GEIANT agents query pre- and post-event Sentinel-2 tiles, run Prithvi classification, cross-reference WxC precipitation data, and produce a verified impact map with full provenance.

Outcome: Actionable, verifiable flood extent in under five minutes. Every affected H3 cell carries confidence score, acquisition timestamp, and Ed25519 agent signature.
27M
Telecom infrastructure
Copper-to-fiber migration as a geospatial operation

National-scale FTTH rollouts span hundreds of thousands of street cabinets, millions of premises, and hard regulatory deadlines. GEIANT gives every installation crew, every cabinet, and every activated line a verifiable H3-anchored identity — turning a logistics problem into a provable audit trail.

Outcome: Every FTTH activation is timestamped, cell-anchored, and signed. Regulator reporting becomes automatic. Deadline compliance is cryptographically provable — not just self-reported.
Reference deployment

From architecture to live infrastructure.

GEIANT is running in production, processing real satellite data over real territory.

GEIANT
M+
Smart meters under continuous agent monitoring
< 0.2%
Typical cloud cover on operational Sentinel-2 tiles
Ed25519
Every agent signed with GNS identity key
100%
Audit coverage — zero unverified escalations
Open source + commercial

The foundation is open. The infrastructure is managed.

GEIANT follows the proven open-core model: a self-hostable open core under Apache 2.0, and a managed platform that handles infrastructure, scaling, and compliance for teams that need it.

OSS

Open source (Apache 2.0)

github.com/GNS-Foundation
  • @geiant/core GeoRouter, base agent types, registry protocol
  • @geiant/mcp-postgis PostGIS MCP server — 11 spatial tools
  • @geiant/mcp-gdal GDAL MCP server — raster processing
  • @geiant/mcp-perception Tile pipeline, Planetary Computer integration
  • GNS Protocol spec — TrIP RFC, Proof-of-Trajectory whitepaper (CC-BY)
  • GEIANT Protocol spec — geospatial agent routing (CC-BY)
IETF
TrIP Internet-Draft submitted to IETF. Trajectory-based Recognition of Identity Protocol — the open standard for physical-world agent identity. GEIANT is built on it.

GEIANT Cloud

Managed platform · Zero ops
  • Routing Managed GeoRouter on Railway — auto-scaling, SLA-backed
  • Perception Dedicated Prithvi-EO-2.0 endpoints — no GPU provisioning
  • Memory Spatial Memory with 12-month retention, queryable via PostGIS MCP
  • Dashboard Live territory map, delegation chains, agent audit log
  • Identity Multi-tenant registry with RBAC and GNS-anchored agent keys
  • Enterprise MSA, dedicated support, 99.9% uptime SLA
Identity layer for every AI framework

GNS-AIP SDK — agent identity without the geospatial stack.

GEIANT is the full physical-world deployment. GNS-AIP is the identity and delegation layer — standalone, open source, and framework-agnostic. Drop it into any LangChain, CrewAI, or OpenAI Agents deployment in under 5 lines.

Every AI agent that runs code, queries data, or takes action needs to answer three questions a regulator will eventually ask: Who authorized this agent? What scope was it given? Can you prove it?

GNS-AIP answers all three with Ed25519 cryptographic identity, signed delegation chains, and a virtual breadcrumb trail — without recording prompts or outputs. The identity primitive is identical to GEIANT's, which means an agent proven in the field can be trusted in the cloud.

→ Full developer docs at gns-aip.gcrumbs.com
GEIANT is GNS-AIP + the physical world. When your agent needs to act on satellite imagery, validate a grid anomaly, or operate within a regulatory territory boundary — that's when the full GEIANT stack comes in. GNS-AIP handles the identity. GEIANT handles everything the physical world adds on top.
5 framework integrations · Apache 2.0
LC
LangChain
TypeScript
OAI
OpenAI Agents
TypeScript
VAI
Vercel AI
TypeScript · Edge
CAI
CrewAI
Python
AG2
AutoGen / AG2
Python
341
Tests passing
CI on every push
TypeScript
Python
LangChain
CrewAI
// npm install @gns-aip/sdk
import { generateAgentIdentity, createDelegationCert, computeComplianceScore } from '@gns-aip/sdk';

// 1. Agent gets a permanent Ed25519 identity
const agent = generateAgentIdentity();

// 2. Human principal delegates scope — cryptographically signed
const cert = createDelegationCert({
  from: humanKeypair,
  to: agent.publicKey,
  scope: ['read', 'write'],
  territory: ['8428309ffffffff'],  // H3 cell — EU jurisdiction
  ttl: 86400
});

// 3. Check compliance tier at any time
const score = computeComplianceScore(agent);
// → { tier: 'VERIFIED', score: 62, dimensions: { delegation: 25, territory: 18 } }
# pip install gns-aip
from gns_aip import AgentIdentity, DelegationCert, ComplianceScorer

# 1. Create agent with Ed25519 keypair
agent = AgentIdentity.generate()

# 2. Delegate from human principal
cert = DelegationCert.create(
    issuer=human_keypair,
    subject=agent.public_key,
    scope=["read", "write"],
    territory=["8428309ffffffff"],
    ttl=86400
)

# 3. Score compliance — real-time, auditable
score = ComplianceScorer.compute(agent)
# ComplianceScore(tier='VERIFIED', score=62)
// npm install langchain-gns-aip
import { GnsAipCallbackHandler } from 'langchain-gns-aip';
import { LLMChain } from 'langchain/chains';

// One callback handler — that's it
const handler = new GnsAipCallbackHandler({ agent, delegation: cert });

const chain = new LLMChain({
  llm,
  prompt,
  callbacks: [handler]   // Every LLM call now carries a signed identity
});

// Every chain invocation is now signed, scoped, and auditable
await chain.invoke({ input: "Analyse this grid anomaly" });
# pip install crewai-gns-aip
from crewai_gns_aip import GnsIdentityTool, gns_step_callback
from crewai import Crew, Agent, Task

analyst = Agent(
    role="Grid Analyst",
    tools=[GnsIdentityTool(identity=agent, delegation=cert)]
)

crew = Crew(
    agents=[analyst],
    tasks=[task],
    step_callback=gns_step_callback   # Every step signed + logged
)

crew.kickoff()  # Full delegation chain in audit log
OSS / Free
$0
The full SDK. Apache 2.0. Self-hosted validation. Use commercially, no restrictions.
  • @gns-aip/sdk — TypeScript + Python
  • All 5 framework integrations
  • Ed25519 identity + delegation chains
  • Local compliance scoring
  • 341 tests · GitHub Actions CI
View on GitHub →
Enterprise
Custom
Unlimited agents, dedicated endpoints, RBAC, custom data residency. EU AI Act reporting included.
  • Unlimited agents + sub-delegations
  • Dedicated validation infrastructure
  • EU AI Act compliance reports
  • SCIM directory integration
  • Dedicated CSM + custom MSA
Talk to us
Pricing

Start open. Scale managed.

Self-host the open source stack for free. Move to GEIANT Cloud when you need managed inference, spatial memory, and enterprise SLAs.

Free / OSS
$0
Self-hosted. Community support. No SLA. Perfect for research and prototypes.
  • @geiant/core, mcp-postgis, mcp-gdal
  • Tile pipeline (Planetary Computer)
  • Your own GPU / HF endpoints
  • Community support via GitHub
  • Apache 2.0 — use commercially
View on GitHub →
Developer
$99/mo
Managed routing and registry. Up to 10 agents. Ideal for pilots and proofs of concept.
  • Up to 10 connected agents
  • 1,000 perception inference calls/mo
  • Managed GeoRouter on Railway
  • Spatial Memory — 3-month retention
  • Email support
Get started
Enterprise
Custom
Dedicated endpoints. Unlimited agents. Custom MSA. Designed for national-scale deployments.
  • Unlimited agents
  • Dedicated HF GPU endpoints
  • Custom retention & data residency
  • Dedicated CSM + Slack
  • 99.9% SLA + custom MSA
Talk to us

Request early access.

GEIANT is in active deployment with enterprise customers. Tell us about your use case and we'll be in touch within 24 hours.

We received your request.

A member of our team will reach out within 24 hours.
In the meantime, check out the open-source repos and the GNS-AIP SDK.

No pitch decks. No sales process. Just a direct conversation about your deployment.