Skip to main content
Glama

pubchem-mcp-server

by io.github.cyanheads
OverviewDiscussionsAdmin

Get Compound Interactions

pubchem_get_compound_interactions
Read-onlyIdempotent

Get a compound's interaction data: drug-food interactions, and the molecular targets its bioassays name — one entry per target as kinds defines it, with its assay counts by outcome and most potent active value, Active targets first, most potent first. Each entry carries its originating source. Results are paged per kind, with the record total and the next offset reported for each. Drug-food coverage is richest for approved drugs; target covers any compound PubChem has assay data for, and pubchem_get_bioactivity lists one target's assays when given the entry's GeneID or accession. Drug-drug interactions (DrugBank) are unavailable while PubChem's SDQ service refuses automated requests: an explicit drug-drug request is reported in failedKinds while the other kinds still return.

Input Schema

NameRequiredDescriptionDefault
cidYesPubChem Compound ID. Resolve from name/SMILES with pubchem_search_compounds.
kindsNoInteraction kinds to fetch. "drug-food" (dietary interactions, from PUG View), "target" (molecular targets from the compound's assay summary: one entry per Gene ID with the accessions the table pairs with it, and one per identifier it never pairs, so a Gene ID and an accession that no row pairs are separate entries even when they name the same protein), "drug-drug" (interactions with other drugs, from PubChem's SDQ service, which refuses automated requests, so it is reported in failedKinds). Default: ["drug-food"].
offsetNoZero-based start position within each requested kind, counted in records (molecular targets, for target) rather than returned entries. The same offset applies to every kind in the call, and the kinds advance at different rates — when paging past the first page, request one kind per call and pass that kind's nextOffset. Default: 0.
maxEntriesNoMax entries per kind per page. The per-call cap is 50: a larger value (up to 1000) reads a page of 50 per kind and is reported in cappedInputs. Well-studied drugs have a long tail of interactions; use offset to reach the ones past this page. Default: 10.

Output Schema

NameRequiredDescriptionDefault
cidNoPubChem Compound ID.
errorNoPresent when the call failed. Absent on success.
noticeNoGuidance when a kind failed, when no interaction data was found, when the offset runs past every requested kind, when target names could not be retrieved, or when further pages remain. Absent when this page is complete and every kind resolved.
offsetNoZero-based start position read within each requested kind.
pagingNoPer-kind page position, one entry per requested kind that was retrieved. A kind listed in failedKinds is absent — its position is unknown, not exhausted.
entriesNoInteraction entries across the requested kinds.
truncatedNoTrue when at least one requested kind has source records remaining past this page. Which kinds, and how many records each holds, is in paging[].truncated / paging[].totalRecords.
nextOffsetNoOffset to pass on the next call, reported when exactly one requested kind has records remaining. Omitted when none do, and when several do — those advance to different positions, so read paging[].nextOffset instead.
failedKindsNoInteraction kinds that could not be retrieved (comma-separated). The returned entries cover the kinds that succeeded. The notice says, per failed kind, whether a retry can help: a timeout or upstream error can clear, while a refusal from PubChem's SDQ service (drug-drug) does not.
cappedInputsNoInputs that asked for more than their per-call cap, with the value requested and the value served. An ID batch lists the IDs past the cap in unprocessed. Omitted when every input was within its cap.
returnedCountNoTotal interaction entries returned across all kinds.
requestedKindsNoInteraction kinds requested (comma-separated).

Schema Changelog

Changes observed during successful MCP inspections.

  1. Changed9 schema fields changed
    • changedInput schema / properties / kinds / default
      Previous value: -[
-  "drug-drug"
-]New value: +[
+  "drug-food"
+]
    • changedInput schema / properties / kinds / description
      Previous value: -"Interaction kinds to fetch. \"drug-drug\" (interactions with other drugs), \"drug-food\" (dietary interactions), \"target\" (binding/activity against molecular targets). \"drug-drug\" and \"target\" come from PubChem's SDQ service, which can refuse automated requests; \"drug-food\" comes from PUG View. Per-target activity is also available from pubchem_get_bioactivity. Default: [\"drug-drug\"]."New value: +"Interaction kinds to fetch. \"drug-food\" (dietary interactions, from PUG View), \"target\" (molecular targets from the compound's assay summary: one entry per Gene ID with the accessions the table pairs with it, and one per identifier it never pairs, so a Gene ID and an accession that no row pairs are separate entries even when they name the same protein), \"drug-drug\" (interactions with other drugs, from PubChem's SDQ service, which refuses automated requests, so it is reported in failedKinds). Default: [\"drug-food\"]."
    • changedInput schema / properties / offset / description
      Previous value: -"Zero-based start position within each requested kind, counted in source records rather than returned entries. The same offset applies to every kind in the call, and the kinds advance at different rates — when paging past the first page, request one kind per call and pass that kind's nextOffset. Default: 0."New value: +"Zero-based start position within each requested kind, counted in records (molecular targets, for target) rather than returned entries. The same offset applies to every kind in the call, and the kinds advance at different rates — when paging past the first page, request one kind per call and pass that kind's nextOffset. Default: 0."
    • changedOutput schema / properties / entries / items / properties / partner / description
      Previous value: -"Interacting compound, food, or target name as the source reports it."New value: +"Interacting compound as the source reports it. For a target: its gene or protein name and organism, then the identifier pubchem_get_bioactivity filters on (GeneID:N → targetGeneId, accession:X → targetAccession); the identifier alone when no name resolves. A target whose identifiers join several Gene IDs is named by the accession they share and says \"group of N Gene IDs\"."
    • changedOutput schema / properties / entries / items / properties / source / description
      Previous value: -"Originating source (e.g. \"DrugBank\", \"BindingDB\")."New value: +"Originating source (e.g. \"DrugBank\", \"PubChem BioAssay\")."
    • changedOutput schema / properties / entries / items / properties / text / description
      Previous value: -"The interaction statement."New value: +"The interaction statement. For a target: the assays naming it by outcome and its most potent Active value with that AID, in µM."
    • changedOutput schema / properties / failedKinds / description
      Previous value: -"Interaction kinds that could not be retrieved (comma-separated). The returned entries cover the kinds that succeeded. The notice says, per failed kind, whether a retry can help: a timeout or upstream error can clear, while a refusal from PubChem's SDQ service (drug-drug, target) does not."New value: +"Interaction kinds that could not be retrieved (comma-separated). The returned entries cover the kinds that succeeded. The notice says, per failed kind, whether a retry can help: a timeout or upstream error can clear, while a refusal from PubChem's SDQ service (drug-drug) does not."
    • changedOutput schema / properties / notice / description
      Previous value: -"Guidance when a kind failed, when no interaction data was found, when the offset runs past every requested kind, or when further pages remain. Absent when this page is complete and every kind resolved."New value: +"Guidance when a kind failed, when no interaction data was found, when the offset runs past every requested kind, when target names could not be retrieved, or when further pages remain. Absent when this page is complete and every kind resolved."
    • changedOutput schema / properties / paging / items / properties / totalRecords / description
      Previous value: -"Source records available for this kind, across all pages. Entries are derived from these records and can be fewer: a \"target\" record naming no molecular target and a \"drug-drug\" record carrying no statement both yield nothing, and duplicate measurements collapse within a page. Pages divide the records, not the entries, so a duplicate split across two pages is reported on both."New value: +"Records available for this kind, across all pages: interaction records for drug-drug and drug-food, molecular targets for target. Entries are derived from these records and can be fewer: a drug-drug record carrying no statement yields nothing. Pages divide the records, not the entries."
  2. Changed6 schema fields changed
    • changedInput schema / properties / cid / maximum
      Previous value: -9007199254740991New value: +2147483647
    • changedInput schema / properties / kinds / description
      Previous value: -"Interaction kinds to fetch. \"drug-drug\" (interactions with other drugs), \"drug-food\" (dietary interactions), \"target\" (binding/activity against molecular targets). Default: [\"drug-drug\"]."New value: +"Interaction kinds to fetch. \"drug-drug\" (interactions with other drugs), \"drug-food\" (dietary interactions), \"target\" (binding/activity against molecular targets). \"drug-drug\" and \"target\" come from PubChem's SDQ service, which can refuse automated requests; \"drug-food\" comes from PUG View. Per-target activity is also available from pubchem_get_bioactivity. Default: [\"drug-drug\"]."
    • changedInput schema / properties / maxEntries / description
      Previous value: -"Max entries per kind per page (1-50). Well-studied drugs have a long tail of interactions; use offset to reach the ones past this page. Default: 10."New value: +"Max entries per kind per page. The per-call cap is 50: a larger value (up to 1000) reads a page of 50 per kind and is reported in cappedInputs. Well-studied drugs have a long tail of interactions; use offset to reach the ones past this page. Default: 10."
    • changedInput schema / properties / maxEntries / maximum
      Previous value: -50New value: +1000
    • addedOutput schema / properties / cappedInputs
      Added value: +{
+  "description": "Inputs that asked for more than their per-call cap, with the value requested and the value served. An ID batch lists the IDs past the cap in unprocessed. Omitted when every input was within its cap.",
+  "items": {
+    "additionalProperties": false,
+    "description": "One input served at its per-call cap.",
+    "properties": {
+      "input": {
+        "description": "Input name, e.g. \"maxResults\" or \"cids\".",
+        "type": "string"
+      },
+      "requested": {
+        "description": "Value requested: the page size, or the number of IDs in the batch.",
+        "type": "number"
+      },
+      "served": {
+        "description": "Value served: the per-call cap.",
+        "type": "number"
+      },
+      "unprocessed": {
+        "description": "ID batches only: IDs past the cap, listed in unprocessed in input order.",
+        "type": "number"
+      }
+    },
+    "required": [
+      "input",
+      "requested",
+      "served"
+    ],
+    "type": "object"
+  },
+  "type": "array"
+}
    • changedOutput schema / properties / failedKinds / description
      Previous value: -"Interaction kinds that could not be retrieved (comma-separated). The returned entries cover the kinds that succeeded; retry to re-attempt the failed ones."New value: +"Interaction kinds that could not be retrieved (comma-separated). The returned entries cover the kinds that succeeded. The notice says, per failed kind, whether a retry can help: a timeout or upstream error can clear, while a refusal from PubChem's SDQ service (drug-drug, target) does not."
  3. Changed2 schema fields changed
    • removedInput schema / properties / cid / exclusiveMinimum
      Removed value: -0
    • addedInput schema / properties / cid / minimum
      Added value: +1
  4. Changed7 schema fields changed
    • changedInput schema / $schema
      Previous value: -"http://json-schema.org/draft-07/schema#"New value: +"https://json-schema.org/draft/2020-12/schema"
    • addedInput schema / additionalProperties
      Added value: +false
    • changedOutput schema / $schema
      Previous value: -"http://json-schema.org/draft-07/schema#"New value: +"https://json-schema.org/draft/2020-12/schema"
    • addedOutput schema / anyOf
      Added value: +[
+  {
+    "not": {
+      "required": [
+        "error"
+      ]
+    },
+    "required": [
+      "cid",
+      "entries",
+      "paging",
+      "requestedKinds",
+      "returnedCount",
+      "truncated",
+      "offset"
+    ]
+  },
+  {
+    "required": [
+      "error"
+    ]
+  }
+]
    • addedOutput schema / properties / error
      Added value: +{
+  "additionalProperties": {},
+  "description": "Present when the call failed. Absent on success.",
+  "properties": {
+    "code": {
+      "description": "JSON-RPC error code for this failure.",
+      "maximum": 9007199254740991,
+      "minimum": -9007199254740991,
+      "type": "integer"
+    },
+    "data": {
+      "additionalProperties": {},
+      "properties": {
+        "reason": {
+          "description": "Machine-readable failure mode.",
+          "type": "string"
+        },
+        "recovery": {
+          "additionalProperties": {},
+          "description": "Actionable next step for the caller.",
+          "properties": {
+            "hint": {
+              "type": "string"
+            }
+          },
+          "required": [
+            "hint"
+          ],
+          "type": "object"
+        },
+        "retryable": {
+          "description": "Whether retrying may succeed.",
+          "type": "boolean"
+        }
+      },
+      "type": "object"
+    },
+    "message": {
+      "description": "Human-readable description of what went wrong.",
+      "type": "string"
+    }
+  },
+  "required": [
+    "code",
+    "message"
+  ],
+  "type": "object"
+}
    • addedOutput schema / properties / truncated
      Added value: +{
+  "description": "True when at least one requested kind has source records remaining past this page. Which kinds, and how many records each holds, is in paging[].truncated / paging[].totalRecords.",
+  "type": "boolean"
+}
    • removedOutput schema / required
      Removed value: -[
-  "cid",
-  "entries",
-  "paging",
-  "requestedKinds",
-  "returnedCount",
-  "offset"
-]
  5. Changed7 schema fields changed
    • changedInput schema / properties / maxEntries / description
      Previous value: -"Max entries per kind (1-50). Well-studied drugs have a long tail of interactions. Default: 10."New value: +"Max entries per kind per page (1-50). Well-studied drugs have a long tail of interactions; use offset to reach the ones past this page. Default: 10."
    • addedInput schema / properties / offset
      Added value: +{
+  "default": 0,
+  "description": "Zero-based start position within each requested kind, counted in source records rather than returned entries. The same offset applies to every kind in the call, and the kinds advance at different rates — when paging past the first page, request one kind per call and pass that kind's nextOffset. Default: 0.",
+  "maximum": 2147483646,
+  "minimum": 0,
+  "type": "integer"
+}
    • addedOutput schema / properties / nextOffset
      Added value: +{
+  "description": "Offset to pass on the next call, reported when exactly one requested kind has records remaining. Omitted when none do, and when several do — those advance to different positions, so read paging[].nextOffset instead.",
+  "type": "number"
+}
    • changedOutput schema / properties / notice / description
      Previous value: -"Guidance when no interaction data was found for the requested kinds."New value: +"Guidance when a kind failed, when no interaction data was found, when the offset runs past every requested kind, or when further pages remain. Absent when this page is complete and every kind resolved."
    • addedOutput schema / properties / offset
      Added value: +{
+  "description": "Zero-based start position read within each requested kind.",
+  "type": "number"
+}
    • addedOutput schema / properties / paging
      Added value: +{
+  "description": "Per-kind page position, one entry per requested kind that was retrieved. A kind listed in failedKinds is absent — its position is unknown, not exhausted.",
+  "items": {
+    "additionalProperties": false,
+    "description": "Page position for one interaction kind.",
+    "properties": {
+      "kind": {
+        "description": "Interaction category this page covers.",
+        "enum": [
+          "drug-drug",
+          "drug-food",
+          "target"
+        ],
+        "type": "string"
+      },
+      "nextOffset": {
+        "description": "Offset to pass on the next call to continue this kind past the current page. Omitted when no records remain.",
+        "type": "number"
+      },
+      "returnedCount": {
+        "description": "Interaction entries returned for this kind.",
+        "type": "number"
+      },
+      "totalRecords": {
+        "description": "Source records available for this kind, across all pages. Entries are derived from these records and can be fewer: a \"target\" record naming no molecular target and a \"drug-drug\" record carrying no statement both yield nothing, and duplicate measurements collapse within a page. Pages divide the records, not the entries, so a duplicate split across two pages is reported on both.",
+        "type": "number"
+      },
+      "truncated": {
+        "description": "True when source records remain for this kind past the current page.",
+        "type": "boolean"
+      }
+    },
+    "required": [
+      "kind",
+      "returnedCount",
+      "totalRecords",
+      "truncated"
+    ],
+    "type": "object"
+  },
+  "type": "array"
+}
    • changedOutput schema / required
      Previous value: -[
-  "cid",
-  "entries",
-  "requestedKinds",
-  "returnedCount"
-]New value: +[
+  "cid",
+  "entries",
+  "paging",
+  "requestedKinds",
+  "returnedCount",
+  "offset"
+]
  6. Changed1 schema field changed
    • removedOutput schema / properties / entries / items / properties / severity
      Removed value: -{
-  "description": "Raw severity as the source reports it — not normalized across sources, and frequently unset (most sources embed severity in the statement text).",
-  "type": "string"
-}
  7. Added

TDQS

A4.8/5.0
Behavior5/5

Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?

Goes well beyond the readOnly/idempotent/openWorld annotations: discloses ordering (active targets first, most potent first), that results are paged per kind with record totals and next offsets, that failedKinds reports unavailable drug-drug requests while others succeed, and that maxEntries above 50 is capped and echoed in cappedInputs. This is exactly the kind of failure-mode and output-shape context annotations cannot carry.

Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.

Conciseness4/5

Is the description appropriately sized, front-loaded, and free of redundancy?

Front-loaded with what is returned and in what order, then paging, then the drug-drug caveat, then the sibling pointer. Some sentences are dense and clause-stacked ('one entry per target as kinds defines it, with its assay counts by outcome and most potent active value'), costing a little readability, but no sentence is filler.

Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.

Completeness5/5

Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?

For a paged, multi-kind read tool with an output schema and full annotations, the description covers everything an agent needs to call it correctly: valid kinds, default kind, offset units, the 50-entry cap, ordering, and the degraded drug-drug path. Return-value structure is correctly left to the output schema.

Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.

Parameters4/5

Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?

With 100% schema coverage the baseline is 3, but the description adds real meaning: offset is counted in records rather than returned entries and applies to every kind simultaneously, so kinds advance at different rates and past the first page you should request one kind per call. That nuance about cross-kind offset semantics is not obvious from the schema alone.

Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.

Purpose5/5

Does the description clearly state what the tool does and how it differs from similar tools?

Names a specific verb and resource ('Get a compound's interaction data') and enumerates the concrete kinds returned: drug-food interactions and molecular targets. It explicitly distinguishes itself from the sibling pubchem_get_bioactivity by scope (one target's assays vs. all targets for a compound).

Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.

Usage Guidelines5/5

Does the description explain when to use this tool, when not to, or what alternatives exist?

States when each kind applies ('Drug-food coverage is richest for approved drugs; target covers any compound PubChem has assay data for') and routes to the alternative tool with the exact inputs needed ('pubchem_get_bioactivity lists one target's assays when given the entry's GeneID or accession'). It also tells the agent how to behave when drug-drug is requested (reported in failedKinds, other kinds still return) and advises one kind per call when paging.

Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.

All pubchem-mcp-server toolsInspect
Try in Browser
One endpoint for every MCP toolRayrun runs every MCP server behind one URL and holds the credentials. Each call is decided per client, tool, and argument, then recorded.ray.runAd

Glama MCP Gateway

Add one secure layer between your agents and this server.

Resources