Digital Ocean Systems
Atlantis CLI & Model Context Protocol (MCP) Server
Bridging autonomous AI workflows with empirical ocean science requires open, secure tool-calling standards. Coralfil Labs develops the Atlantis CLI and dedicated Model Context Protocol (MCP) servers to allow external AI models and developer environments to query live oceanographic data, execute hydrodynamic scripts, and verify statutory compliance directly.
Developer Tooling Overview
The Atlantis command line interface provides researchers, marine engineers, and field technicians with terminal access to sovereign marine intelligence. Built with Python and Click, the tool operates entirely offline when deployed aboard maritime vessels or connects to regional networks for real-time telemetry syncing.
Installation & Setup
The CLI is packaged as an open development package. Install directly from the Coralfil OS repository:
# Clone the sovereign repository and install editable package
git clone https://github.com/coralfil/os.git
cd os/atlantis_cli
pip install -e .
# Verify installation and system prerequisites
atlantis audit
Command Line Reference
The table below documents primary command groups and execution options:
atlantis agent-sea
atlantis agent-sea [OPTIONS]Dispatches the Zephyr Master Conductor across the 30-agent canonical roster and 5 oceanic depth divisions under the 6-stage helicinid lifecycle.
--mission, -mMission identifier (default: 'Salish-Sea-Survey')--domain, -dTarget domain specialty (default: 'Ocean-Telemetry-Modeling')--memory-sync / --no-memory-syncSynchronize findings with canonical Hexactine lattice state (default: True)
atlantis liveocean
atlantis liveocean [OPTIONS]Ingests real-time Pacific buoy observations across 16 stations (Hakai, ONC, NOAA, DFO) and processes Sentinel-3 satellite multispectral rasters.
--satellite-tile, -sSentinel-3 scene ID for TerraTorch optical and thermal ingestion--list-tiles, -lList cataloged Sentinel-3 multispectral rasters--region, -rFilter scenes by region (salish_sea, hecate_strait, juan_de_fuca)
atlantis code
atlantis code [OPTIONS]Autonomous software engineering engine (DeepCoder). Features dual-engine speculative drafting (110+ tok/s), hierarchical AST indexing, and unified diff verification.
--prompt, -pRefactoring instruction or coding prompt--target-file, -fPath to specific file to inspect and refactor--dry-run / --applyPreview verified unified diff without writing to disk (default: dry-run)--hexactine-sync / --no-hexactine-syncSynchronize refactoring spicule with Hexactine lattice
atlantis hexactine
atlantis hexactine <COMMAND> [OPTIONS]Inspects, queries, and prunes the degree-6 Hexactine graph memory lattice (+x, -x, +y, -y, +z, -z coordinate axes).
statusDisplays spicule counts, cross-basin connections, and lattice healthquery --spicule-id <ID>Retrieves connected memory nodes, historical baselines, and anomaly linksattach --source <ID> --target <ID>Forms an orthogonal semantic connection between spiculesprune --older-than <DAYS>Cleans uncorroborated ephemeral trace hypotheses
atlantis audit
atlantis auditRuns a full system audit checking GPU acceleration, PyTorch/CUDA versions, local quantization runtime, and verified endpoint connectivity.
--verbose, -vDisplay full hardware register states and endpoint latency histograms
atlantis serve
atlantis serve [OPTIONS]Launches an air-gapped, sovereign edge inference server hosting quantized model checkpoints (AWQ INT4 GEMM / GGUF) for vessel or farm use.
--port, -pLocal port to bind HTTP server (default: 8000)--model, -mModel checkpoint identifier (default: 'Atlantis-Obelisk-v2-AWQ')
atlantis arc
atlantis arc [OPTIONS]Autonomous tripartite solver for ARC-AGI abstract reasoning puzzles, combining hypothesis generation, DSL code synthesis, and epistemic verification.
--puzzle, -pARC task puzzle identifier (default: 'arc_color_symmetry_01')--list-puzzles, -lCatalog of available ARC benchmark puzzles
atlantis benchmark
atlantis benchmark [OPTIONS]Executes evaluation suites against the AboveBoard 52 benchmark, CASMI Spec2Mol inverse discovery, or AboveBoard Marine 52 telemetry sets.
--suite, -sBenchmark suite name ('aboveboard-52', 'casmi-spec2mol', 'marine-52')--output, -oPath to export JSON-formatted evaluation results
Terminal Execution Examples
Below are real execution outputs from the Atlantis command line interface:
$ atlantis agent-sea --mission Salish-Sea-Survey --domain Ocean-Telemetry-Modeling
[Atlantis Agent-Sea] Dispatching mission 'Salish-Sea-Survey' under Zephyr Conductor...
Division: BATHYPELAGIC
Squad: VELL, DECO, ECCO, PERRO, SAMUDRA
On-Demand Skills Mounted: coral-evolution-engine, chemistry-solver, samudra-agent
Task-Shape Route: Multi-Scale Hydrodynamics -> Atlantis-Pyramid (32B)
Lifecycle: 6/6 Stages Verified (TEOS-10 & DFO Compliant)
Hexactine Spicule: SPICULE-PACIFIC-SALISH-495 (Synced to Canonical Lattice)
$ atlantis liveocean --satellite-tile S3A_OL_1_EFR____20261028T184500
[Atlantis LiveOcean] Querying Pacific in-situ telemetry stations...
Station: Hakai Institute Quadra Island (5m depth)
Temp: 10.4 C | Salinity: 29.8 PSU | DO: 7.8 mg/L
Station: ONC Folger Passage Deep (100m depth)
Temp: 8.2 C | Salinity: 33.1 PSU | DO: 4.1 mg/L
[Atlantis LiveOcean] Ingesting Sentinel-3 raster: S3A_OL_1_EFR____20261028T184500...
Chlorophyll-a: 3.42 mg/m3
SST Skin: 11.2 C
Optical Limit: 14.5m
Hexactine Attachment: SPICULE-SATELLITE-S3-20261028 (Attached)
The Atlantis Model Context Protocol (MCP) Server
The Model Context Protocol (MCP) provides an open, sandboxed protocol for connecting AI systems to external data sources and tools. The Atlantis MCP server exposes standardized tool endpoints to agent environments such as Anthropic Claude Code, Google Antigravity, and OpenAI Swarm:

Nautilus MCP Integration Bridge
Standardized tool adapter exposing subsea CTD instruments, regional tide tables, and statutory compliance checkers to frontier agent harnesses with zero sandbox escape risks.
get_buoy_telemetryTool EndpointRetrieves calibrated observations from regional Hakai Institute and Ocean Networks Canada buoys.
compute_aragonite_saturationTool EndpointCalculates localized aragonite saturation states using TEOS-10 thermodynamic equations.
audit_fisheries_complianceTool EndpointValidates planned coastal actions against statutory regulations and live spatial boundaries.
query_hexactine_latticeTool EndpointExplores the 6-layer memory graph to correlate telemetry anomalies with historical observations.
Open Source Repositories
Developer tooling is published openly on GitHub under the Coralfil organization: