ASTRA MCP Server
Server Configuration
Describes the environment variables required to run the server.
| Name | Required | Description | Default |
|---|---|---|---|
| ASTRA_SSE_PORT | No | SSE transport port | 9002 |
| ASTRA_HTTP_PORT | No | Streamable HTTP port | 9003 |
| ASTRA_LOG_LEVEL | No | Log level: debug, info, warn, error | info |
| ASTRA_CORS_ORIGIN | No | CORS allowed origin | * |
Instructions
Guidance the server publishes about itself, which clients place ahead of the tool catalog so the model reads it before choosing anything.
This server publishes no instructions, or was last inspected before Glama recorded them.
Capabilities
Features and capabilities supported by this server
Protocol revision2025-11-25
| Capability | Details |
|---|---|
| tools | {
"listChanged": true
} |
| prompts | {
"listChanged": true
} |
| resources | {
"listChanged": true
} |
Tools
Functions exposed to the LLM to take actions
| Name | Description |
|---|---|
| get_system_statusC | ASTRA System Status |
| get_metricsC | Real-time Metrics |
| get_snn_stateC | SNN Engine State |
| snn_stepC | Advance SNN Simulation |
| snn_resetB | Reset SNN Engine |
| inject_spikesD | Spike Injection |
| get_acm_scoreC | Consciousness Assessment (Proxy) |
| check_ethicsC | IRB Neural Welfare Check |
| set_parameterD | Modify State Parameter |
| get_platform_statusD | Bio-Computing Platforms |
| export_snapshotC | Full State Snapshot |
| simulation_controlD | Simulation Control |
| wm_encodeB | Encode SNN State to Latent Space |
| wm_predictC | Predict Next SNN State in Latent Space |
| wm_planC | CEM Planning for Optimal Spike Injection |
| wm_surpriseC | Violation-of-Expectation Detection |
| wm_train_stepD | Online World Model Training Step |
| wm_statusC | World Model Status & Metrics |
| sensor_visualC | V-JEPA 2 Visual Encoding (Image/Video) |
| sensor_audioB | A-JEPA Audio Encoding (Waveform → Mel → Latent) |
| sensor_olfactoryC | Koniku Kore Olfactory Encoding (Chemoreceptor → Latent) |
| sensor_fuseD | Cross-Modal Attention Fusion |
| sensor_processC | Full Multimodal Pipeline (All Modalities → Fused z) |
| sensor_statusC | Multimodal Sensor Pipeline Status |
Prompts
Interactive templates invoked by user choice
| Name | Description |
|---|---|
| wm-experiment | World Model experiment: encode → predict → compare → plan |
| multimodal-experiment | Full multimodal sensor experiment: visual + audio + olfactory → fused → WM |
| system-health-report | Comprehensive system health report |
| snn-experiment | Controlled SNN experiment |
| ethics-stress-test | Progressive biomarker degradation |
Resources
Contextual data attached and managed by the client
| Name | Description |
|---|---|
| wm-latent | Current latent space state and embedding history |
| wm-predictions | World Model prediction history and accuracy |
| sensors-state | Multimodal sensor pipeline state and last fusion |
| metrics-realtime | Live metrics |
| snn-topology | SNN network architecture |
| acm-state | Consciousness proxy assessment |
| ethics-welfare | IRB compliance report |
| snapshot-current | Complete state dump |
TDQS
Scored across 24 tools
Many status/metrics tools overlap conceptually: get_system_status, get_metrics, get_snn_state, get_platform_status, wm_status, and sensor_status all report some form of state or health. Additionally, sensor_fuse and sensor_process have borderline responsibilities, though wm_* and sensor_* prefixes help distinguish the two main subsystems.
The naming is organized by subsystem prefixes like get_*, wm_*, and sensor_*, but the conventions are mixed: snn_step and snn_reset are command-like, simulation_control is noun-like, and inject_spikes is verb_noun. There is no single consistent pattern across the full tool set, though the prefix grouping keeps it readable.
24 tools sits in the heavy range and requires agents to navigate several distinct subsystems: core SNN simulation, world model, sensors, status, and ethics. The count is defensible given the breadth of the platform, but it is borderline and each tool needs to justify its place.
The tool surface covers a broad lifecycle: SNN stepping/reset, spike injection, parameter changes, snapshots, world model training/prediction/planning, and multimodal sensor processing. Minor gaps exist, such as no explicit SNN parameter retrieval or sensor data ingestion control, but agents can mostly achieve the platform's apparent goals without dead ends.