Flagship Release · v1.2.3

ForgeGIS

GPU-Accelerated Geospatial Compute for the Agent Era

As of May 2026, the only commercially-supported GPU-accelerated geospatial MCP server we have been able to identify. Distributed as pure-Java Maven artifacts — no first-party native code, no GDAL native bindings, no separate native install step. 239 user-facing operations across 11 geospatial categories; 237 callable by AI agents through a native Model Context Protocol server.

239 / 11
user-facing operations across 11 geospatial categories
237 / 239
operations callable by AI agents via the native MCP server
9×–56×
multi-stage pipeline speedup vs GDAL CLI on equivalent workloads

What it is, why it matters, how it's built differently

Three short answers for the technical evaluator.

What it is

ForgeGIS is a GPU-accelerated geospatial compute library and a native Model Context Protocol server, both written in Java. It runs raster, hydrology, visibility, spectral, ML, and point-cloud workloads on the GPU through Java-side compute kernels, alongside a large vector and topology surface in pure Java — and exposes 237 of its 239 operations to AI agents through MCP.

Why it matters

Serious geospatial work has historically meant assembling Python, C++, and native libraries — GDAL, JTS, PostGIS, CUDA — and accepting either CPU-bound throughput or a build pipeline tied to native dependencies. ForgeGIS removes both costs: GPU performance without leaving the JVM, and agent-callable surface without an integration layer.

How it's built differently

Pure-Java Maven distribution. No first-party native code, no GDAL native bindings, no separate native install step. GPU dispatch is via the JOCL OpenCL bindings (JNI bridge bundled in the JOCL JAR, extracted at runtime); the system OpenCL runtime is standard system software bundled with GPU drivers. Deployment is one JAR.

Dual-surface architecture

One library, two surfaces — one designed for agent composition, one for typed single calls.

The MCP server (forgegis-mcp) exposes ForgeGIS through two complementary surfaces, so an AI agent can either compose a multi-stage workflow in a single call or invoke a specific operation directly with typed inputs.

Surface 1 — Pipeline DSL
152 operations

Composable JSON pipeline format. Agents assemble multi-stage workflows (read → reproject → mask → classify → write) inside a single MCP call — non-linear DAGs as of 1.1.

Surface 2 — Typed Intent Tools
60 single-call tools

Strongly-typed MCP tools for the most common operations. One MCP tool per intent, with structured inputs and validated arguments.

68 MCP tools total — 60 typed intent tools, 5 primitive handle/session tools, and 3 pipeline-escape tools · 237 of 239 catalog operations reachable from an agent. New in 1.1: a MINIMAL 8-tool surface (the 5 primitive + 3 escape tools) is advertised by default; opt into the full 68 with --profile standard.

New in ForgeGIS 1.1

Release 1.1 adds 19 user-facing operations over 1.0 — concentrated in the vector and raster→vector surface — and extends the architecture in five ways.

Pipeline

Non-linear DAG engine

The GPU-resident pipeline extends from linear chains to arbitrary directed acyclic graphs. Three executors — raster, vector, and mixed — run branching workflows where a stage fans out to multiple consumers and multiple inputs converge on one operation, with raster intermediates staying resident on the device across the graph. A SavedPipeline v2 wire format carries typed slot references, conditionals, and nested pipelines.

Migration

QGIS model import

A dedicated module reads QGIS .model3 processing models and compiles them to ForgeGIS DAGs, with a compatibility-report mode that names any unsupported algorithm up front. Existing QGIS investments run on the GPU-resident engine without a rebuild — the expression layer is a clean-room QGIS Processing expression interpreter.

Vector surface

Boolean overlay, dissolve, Voronoi & more

The vector surface more than doubled: boolean overlay (difference, intersection), dissolve, Voronoi polygons, variable buffering, field calculators, and a clean-room expression interpreter — closing the full raster↔vector handoff grid.

Vertical datum

Bundled EGM96 geoid

The first dataset ForgeGIS ships. Orthometric/ellipsoidal vertical-datum conversion works out of the box — relevant wherever elevation reference frames matter to the product.

Agent context

Profile-gated discovery

The MCP server advertises a MINIMAL 8-tool surface by default (5 primitive + 3 pipeline-escape), keeping an agent's tool list inside context and rate-limit budgets. Opt into the full 68-tool STANDARD surface with --profile standard; hidden tools remain callable by name — the profile governs discovery, not capability.

New in 1.2.3: the digital twin

GPU simulators exist; what a running ForgeGIS twin adds is the ability to branch one mid-event, and a recording that proves what you were shown. Demonstrated over real SRTM terrain around Hoover Dam: a routed release paused at epoch 54 and forked at ten times the rate, opening a ≈1.3 km³ wedge between the two worlds — exactly what the closed-form law predicts.

Pause a running simulation

State advances in deterministic epochs on the GPU. A run can be paused mid-event and inspected rather than restarted from the beginning.

Fork it, and change one decision

The branch anchors to its parent instead of copying it, so it inherits every epoch of history. Branches differ by parameters only — everything they disagree about afterwards is attributable to the value you overrode.

Check the difference

One conserved ledger across the whole corridor means the gap between two worlds is a number you can test against a closed-form prediction, not a picture you have to trust.

What the digital twin is Hoover Dam showcase

Performance highlights

Benchmarked against GDAL CLI on equivalent workloads. Numbers are summary — full methodology, dataset descriptions, and per-operation timings live in the Technical Brief.

Full methodology, hardware configuration, and per-operation breakdowns: ForgeGIS Technical Brief (PDF).

Vector at scale

The figures above are the GPU raster surface. The vector and topology surface is pure Java and runs on the CPU — so we measured it the same way, on real data, and published where it stops.

A million real footprints

Microsoft building footprints for Houston through a seven-operation chain — reproject, shape metrics, setback buffer, points-in-polygons join, classify — at 124,785 features/sec measured around each operation.

Near-flat across an order of magnitude

Throughput varies by just 2.03× between 100k and 1M features, because every intermediate stays in process as an immutable handle and there is no per-stage file I/O term to grow with N.

A ceiling that says so

The 5M tier is refused before a byte is read, with the heap arithmetic in the message. A scale curve that hides where it stops is a marketing curve, so this one states it.

See the vector scale study

Performance demonstration

62 operations on real terrain — Cop30 DEM (Grand Canyon, Alps, UTM zone 33N) and a 6-band Sentinel-2 scene — side-by-side with GDAL. Wall-clock on both sides, cumulative speedup at the bottom strip, no cuts.

Methodology — biases built in against ForgeGIS

  • ForgeGIS runs on a cold file cache; GDAL runs second on a warmed cache.
  • ForgeGIS kernel compilation cost is included — kernels precompiled at startup, matching production deployment.
  • GDAL is invoked through native SWIG JNI bindings for single operations and through CLI subprocesses for multi-stage pipelines, matching the way real GDAL workflows run.
  • gdal_calc.py invocations include Python interpreter startup × N invocations, as a real GDAL user would pay.

NVIDIA RTX 5070 Laptop GPU · Intel Core 7 240H · Windows 11 · Java 17 · ForgeGIS 1.1.0 · GDAL 3.12.1. Results specific to this stack.

Per-operation JSONL timings from this recording, and the reproducible benchmark harness, available to vetted buyers under NDA — rich@seaglassfoundry.com.

Cop30 DEM contains modified Copernicus DEM data © DLR e.V. 2010–2014, Airbus DS GmbH 2014–2018, provided under COPERNICUS by the European Union and ESA. Sentinel-2 L2A contains modified Copernicus Sentinel data, ESA. GDAL is open-source software developed by the GDAL/OGR Project under an MIT-style licence.

Who it's for

Three audiences, three one-pagers. Each leads with the parts of ForgeGIS that matter most to that reader.

Capability catalog

239 operations across 11 categories. The full catalog with operation-level descriptions is in the Technical Brief.

Terrain14
Bathymetry / Ocean14
Hydrology7
Visibility / RF8
Spectral / ML30
Spatial Geometry94
Filtering / Focal / Morphology14
Raster Mgmt / Vectorize / IO44
Routing7
Point Cloud5
Temporal2
Topology completeness. The Spatial Geometry category includes full DE-9IM coverage: the 8 OGC named predicates, 3 JTS extensions, the Relate operation, and the raw 9-character intersection matrix per geometry-type pair. Few GPU libraries surface the full DE-9IM relate matrix — most stop at named predicates.

Downloads

All ForgeGIS collateral. Direct download — no email gate, no form wall.

Talk to us about ForgeGIS

Licensing questions, evaluation access, partnership inquiries — we read every email.

rich@seaglassfoundry.com