Provides access to the NASA Planetary Data System (PDS) Registry API for searching and exploring missions, celestial bodies, spacecraft, instruments, data collections, and product relationships.
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., "@NASA PDS Registry MCP Server Search for Mars rover Curiosity data products"
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.
An MCP server and reproducible evaluation of multistep natural-language search over the NASA Planetary Data System (PDS) Registry -- the digital data archive for all of NASA's planetary missions, flight and ground-based observations, and laboratory experiments from the 1960s to the present. With more than 1.85 petabytes from 70+ missions, 4,500 datasets, and 700 instruments all peer-reviewed by JPL and other research institutions.
Evaluation study · Results · Dataset · Data engine · Benchmark and reproducibility · MCP server · Development guide
https://github.com/user-attachments/assets/1d6b7035-c07a-4ca3-8bd2-9c19d68c3d5c
The NASA Planetary Data System (PDS) serves as a high-quality, hand-curated data source for the whole community of planetary research. However, searching the PDS Registry is difficult, requiring knowledge of exact PDS4 ontology fields and context identifiers to manually construct queries for its public search APIs. The research chain of scientific question -> experimentation -> findings/results is thus bottlenecked by current PDS search capabilities.
We introduce PDS-MCP, an agentic search interface on top of the NASA PDS Registry. From natural language questions, our agent can iteratively resolve PDS entities, inspect relationships, construct valid filters, and retrieve matching products autonomously constructing PDS API queries with no manual human intervention. We evaluate this on our 300-question benchmark, constructed from our scalable data engine, with results achieving 98.02% macro F1 and 95.67% exact result-set match on our full system, compared with 12.29% and 8.33% for single-call MCP and 0.33% on both metrics without PDS access, demonstrating that it can reliably bridge natural-language questions and structured PDS retrieval.
By releasing PDS-MCP publically, we aim to support the researchers of the Planetary Data Science community by making it easier to access NASA PDS data. Our goal is to provide enhanced search capabilities that enable more effective data exploration and improve accessibility for future research endeavors.
Related MCP server: NASA MCP Server
Can a language-model agent reliably translate natural-language planetary-data requests into complete PDS Registry result sets, and does multistep MCP search outperform the same model with one or zero PDS tool calls?
The benchmark dataset contains 300 independent questions:
Dimension | Coverage |
Multihop questions with two or three context constraints | 225 |
Single-constraint controls | 75 |
Archive-collection retrieval questions | 294 |
Context lookups | 6 |
Target result-set size | 1–98 PDS identifiers |
The complete benchmark data engine and its dataset documentation are included in this repository.
Ground truth was constructed as follows:
Select compatible PDS context objects → Build a PDS API query → Compile and execute it against the live PDS API → Save search results as target → Derive a natural-language questions as inputs
For each benchmark item,
input: the natural-language question
target: the desired search results (from the intermediary PDS query)
Thus, the job of the agent run on the benchmark is to use the NLQ to build a PDS API query (without knowledge of the original one used in the data engine) and get the target search results.
Condition | Model | PDS access | Purpose |
Baseline | GPT-5.6 Luna | None | The same LLM without PDS-MCP or access to the PDS search interface. |
Single-call MCP | GPT-5.6 Luna | At most one live MCP call | Tests whether tool availability without iterative search is sufficient. |
Multistep MCP (ours) | GPT-5.6 Luna | Iterative live MCP calls | Tests entity resolution, refinement, and collection retrieval. |
Condition | Macro precision | Macro recall | Macro F1 | Exact match | Mean calls | Mean latency |
Baseline | 0.33% | 0.33% | 0.33% | 0.33% | 0.00 | 8.00 s |
Single-call MCP | 13.62% | 11.71% | 12.29% | 8.33% | 1.01 | 29.86 s |
Multistep MCP (ours) | 98.18% | 97.97% | 98.02% | 95.67% | 5.28 | 47.24 s |
A request such as:
Find PDS collections from the New Horizons Kuiper Belt Extended Mission 1, collected by the Radio Science Experiment, targeting Arrokoth.
requires the agent to resolve several ontology objects before retrieving data:
Resolve the KEM1 investigation identifier.
Resolve the New Horizons REX instrument identifier.
Resolve Arrokoth and its PDS target identifier.
Search collections using the compatible investigation, instrument, and target constraints.
Return the complete normalized collection-identifier set.
This mirrors the work a user would otherwise perform manually when discovering PDS4 context identifiers and exact Registry filters.
Readers interested in the benchmark can start with the
research/data guide. The underlying
pds_research data engine samples compatible PDS entities,
builds and validates typed query plans, executes live API requests, and writes
the resulting questions and target identifier sets.
Artifact | Description |
Dataset documentation and links to its construction pipeline. | |
Questions, typed query plans, API requests, ground-truth identifiers, difficulty labels, and provenance. | |
Benchmark-generation, validation, query-compilation, and evaluation code. | |
Raw baseline predictions from the LLM without PDS-MCP or PDS search access. | |
Raw single-call predictions. | |
Raw multistep predictions. | |
Graph-ready aggregate comparison. | |
Machine-readable metrics and resource measurements. | |
Multistep hop-count distribution and exact accuracy by call count. | |
Headline, difficulty, retrieval-error, and trajectory-length figures in PNG and SVG formats. | |
Dependency-free script that regenerates the supporting SVG figures from committed metrics. |
Recompute identifier-set metrics with:
pds-research score \
research/data/main-300-live.jsonl \
research/results/codex-live-main-300-gpt-5.6-luna.jsonlSee DEVELOPMENT.md for environment setup, server execution, MCP client configuration, and test commands.
The FastMCP server in src/pds_mcp_server.py exposes
live PDS Registry operations for:
Investigation, target, instrument-host, and instrument search.
Context-product traversal.
Collection search by investigation, target, instrument, and host.
Detailed product retrieval by PDS identifier.
The server can be used from Claude Desktop, Cursor, Codex, or another MCP-compatible host. Setup instructions are in DEVELOPMENT.md.
Find Apollo 17 collections produced by the Lunar Surface Experiments Package Heat Flow Experiment and targeting the Moon.
Find InSight collections produced by the Auxiliary Payload Sensor Subsystem temperature and wind sensor and targeting Mars.
Find Cassini collections produced by the Imaging Science Subsystem Wide Angle camera and hosted on the Cassini Orbiter.
Find New Horizons KEM1 radio-science collections targeting Arrokoth.
These examples are drawn from the released benchmark; the complete set is
available in main-300-live.jsonl.
Code is released under the MIT License.
PDS Registry API: contact pds-operator@jpl.nasa.gov or open an issue in the
PDS API repository.
This server and study: open an issue in this repository.
Crawl a single PDS Context product and return other PDS Context products it is associated with. Ex. Mars 2020: Perseverance Rover (Investigation) is associated with Mars (Target) and Mastcam (Instrument), so it returns Mars and Mastcam.
WARNING: Takes a long time to run and performs several sequential API calls. Use wisely.
Inputs: urn (required).
Get a single PDS product by its URN identifier.
Inputs: urn (required). Returns: result
Search PDS data collections filtered by instrument, target, instrument host, or investigation. Example: Mars Reconnaissance Orbiter HiRISE data collections targeting Mars.
Inputs: limit, ref_lid_target, ref_lid_instrument, ref_lid_investigation, +1 more. Returns: result
Search PDS Context products that are Instrument Hosts (spacecraft, rovers, telescopes).
Instrument Hosts are platforms that carry scientific instruments: spacecraft, rovers, landers, telescopes. Example: Cassini Orbiter - urn:nasa:pds:context:instrument_host:spacecraft.cassini
Use for queries about specific spacecraft, rovers, or platforms that carry instruments.
Inputs: limit, keywords, instrument_host_type. Returns: result
Search the latest-versioned instances of PDS Context products that are Instruments.
Instruments are scientific devices (cameras, spectrometers, etc.) used on spacecraft to collect data. Example: Cassini RADAR - urn:nasa:pds:context:instrument:radar.cassini
Use for queries about specific instruments, instrument types, or instruments on missions/spacecraft.
Inputs: limit, keywords, instrument_type. Returns: result
Search PDS Context products that are Investigations (missions/projects).
Investigations are organized missions or projects that collect scientific data. Example: Cassini-Huygens - urn:nasa:pds:context:investigation:mission.cassini-huygens
Use for queries about space missions, mission timelines, or finding missions that studied specific targets.
Inputs: limit, keywords. Returns: result
Search PDS Context products that are Targets (celestial bodies, phenomena).
Targets are objects of scientific study: planets, moons, asteroids, comets, etc. Example: Mars - urn:nasa:pds:context:target:planet.mars
Use for queries about specific celestial bodies, finding targets by type, or targets studied by missions.
Inputs: limit, keywords, target_type. Returns: result
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
Scored across 7 tools
The four search_* tools map cleanly onto distinct PDS Context product types (Investigations, Targets, Instrument Hosts, Instruments), and descriptions reinforce the boundaries. crawl_context_product, search_collections, and get_product each target clearly different operations, leaving little room for misselection.
All tool names follow a predictable verb_noun snake_case pattern (search_investigations, crawl_context_product, get_product). The search_ prefix is used consistently for the five query tools, and the verb choices accurately reflect each operation.
Seven tools is well-scoped for a specialized registry query server, with each tool earning its place by covering a distinct resource or operation. Nothing feels redundant or missing count-wise.
Core workflows are covered: search across the four context types, search collections, traverse associations, and fetch by URN. Minor gaps exist—no dedicated search_products or bundle search, and the only product retrieval is by exact URN, but agents can largely work around these.
Discover, verify, and access NASA Earth science datasets via the Common Metadata Repository (CMR)
Public tools to understand Dynamik, discover datasets, and connect account-scoped capabilities.
NASA MCP — wraps NASA Open APIs (api.nasa.gov)
Search public Australian environmental evidence with provenance across authoritative catalogues.
