ImmunoGraph MCP
ImmunoGraph MCP is a provenance-aware MCP server with 46 typed tools for inspectable computational epitope research.
Validate/normalize FASTA and generate MHC-I/II peptide windows.
Resolve MHC-I/II and B-cell predictions through configured live/cache/fixture connectors; run synthetic demos and population coverage.
Normalize scores, compute consensus/coverage, rank/optimize candidates and constructs, calibrate confidence, and apply overlap, duplicate, threshold, and constraint rules.
Fetch/validate structures and compounds; map epitopes; calculate accessibility, AlphaFold confidence, and binding pockets; create Mol* viewers.
Validate/deduplicate compounds, calculate descriptors, prepare ligands/receptors, and validate docking boxes.
Run AutoDock Vina docking or fixtures, cluster poses, and extract PLIP protein-ligand interactions.
Generate reports, export candidates/CSVs, visualize results, explain candidates, and export redacted workflow traces.
Describe/run bounded agentic workflows, chat over supplied evidence, and export a checksummed research package with artifacts and provenance.
Results distinguish LIVE, CACHED, SYNTHETIC, and FIXTURE sources; outputs are for computational research, not clinical or experimental validation.
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., "@ImmunoGraph MCP Validate this FASTA sequence and generate 9-mer peptide windows"
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.
ImmunoGraph MCP
A standard Model Context Protocol server for computational epitope research. Call 46 typed tools for sequence processing, immune evidence, structure and chemistry analysis, docking, and reviewable exports. Results carry provenance so a live provider result stays distinct from a fixture or synthetic demonstration.
Problem · Who it is for · User flow · Get started · 3D outcomes · All 46 tools · Contribute
ImmunoGraph supportscomputational research, not clinical or experimental validation. A predicted epitope, docking pose, score, or exported 3D model does not establish binding, vaccine efficacy, or safety. Fixture and synthetic results are demonstrations and are labelled as such.
The problem
An epitope research workflow crosses different kinds of software: sequence processing, immune predictors, HLA coverage, structure databases, molecular preparation, docking, interaction analysis, and reporting. Each step has its own input format, runtime, and scientific assumptions. When an LLM or research application chains them together, it must also keep track of which inputs and methods produced each result. A plausible looking score or 3D view is easy to misread if its source is missing.
ImmunoGraph gives an MCP host one typed interface for these steps. The host can discover and call individual tools, pass outputs between them, and inspect explicit source labels and exported bytes. This reduces custom integration code for a reviewable computational workflow; it does not replace scientific judgment or validate a prediction by itself.
Related MCP server: FlowProof
Who it is for
Audience | Typical use |
Computational immunology and vaccine research teams | Assemble and review sequence, MHC, B-cell, coverage, and structural evidence for candidate exploration |
Structural bioinformatics and medicinal chemistry researchers | Prepare coordinates, run configured docking tools, inspect poses and contacts, and export the actual files |
Developers building LLM research assistants | Expose typed scientific operations through standard MCP instead of writing a separate adapter for every step |
Educators and reproducibility reviewers | Demonstrate a workflow with clearly marked fixtures, then inspect method boundaries, provenance, and artifacts |
The outputs are for research and education. Clinical decisions, efficacy claims, and experimental validation require independent methods and evidence.
What you can do
Explore | Analyze | Review and share |
Validate FASTA, generate peptide windows, retrieve PDB structures and PubChem compounds | Call enabled IEDB predictors; calculate consensus and coverage; run RDKit, FreeSASA, fpocket, Meeko, Vina, and PLIP where installed | Rank caller-supplied evidence; export CSV, PDB, SDF, PDBQT, docking poses, Mol* or docking viewers, and a checksummed research ZIP |
This is one server with 46 tools, not 46 independent live data providers. Each tool is callable on its own. A client passes results between calls when building a workflow.
User flow
The researcher works in an MCP-enabled chat or application. The client chooses the calls and carries evidence forward; ImmunoGraph executes each requested operation and returns a typed result.
Step | Researcher or client action | ImmunoGraph response |
1. Connect | Add the stdio server to an MCP host; enable only the live connectors needed for the study | The host discovers 46 tools and their input schemas |
2. Define inputs | Provide a FASTA sequence and study criteria |
|
3. Gather evidence | Request available MHC-I/II, B-cell, population, or other calculations | Each call returns scores or a structured failure with source status; live predictors require their configured service |
4. Prioritize | Pass selected evidence, thresholds, and coverage data into the scoring and constraint tools | Normalized scores, consensus, exclusions, and ranked candidates depend on the supplied evidence |
5. Inspect structure | Optionally fetch PDB/compound data and run accessibility, pocket, ligand, docking, or contact tools | Coordinate-based outputs, pose files, 3D viewer HTML, and analysis data where the scientific runtime is installed |
6. Review and export | Inspect source labels and parameters, then request reports, CSV, trace, or research ZIP | File bytes with lengths and SHA-256 hashes for independent saving and verification |
This flow is modular: a user can call one tool without running every stage. run_agentic_workflow executes caller-supplied, approved tool calls; without those calls it returns a plan rather than silently launching a full study. See architecture and the tool reality matrix for exact execution boundaries.
Get started
Install the public release
Node.js 20 or newer is required. The versioned tarball is hosted on GitHub and installs with npm; the immunograph-mcp name is not yet published on the npm registry.
npm install --global https://github.com/Ajey95/immuno-graph/releases/download/v0.2.1/immunograph-mcp-0.2.1.tgzRegister the installed command in an MCP host that supports stdio servers:
{
"mcpServers": {
"immunograph": {
"command": "immunograph-mcp"
}
}
}The server speaks MCP on stdio, so launch it through your host. If the host cannot find the globally installed command, configure its absolute path. The GitHub release also includes the source and a SHA-256 checksum for the tarball. The downloaded public tarball matched that checksum.
Run the public container
The official MCP Registry entry points to the public OCI image. For a local stdio host, use:
{
"mcpServers": {
"immunograph": {
"command": "docker",
"args": [
"run", "--rm", "-i",
"ghcr.io/ajey95/immuno-graph:0.2.1",
"node", "dist/index.js"
]
}
}
}The container also supports guarded Streamable HTTP at /mcp. See HTTP deployment for its required security settings.
Build from source
git clone https://github.com/Ajey95/immuno-graph.git
cd immuno-graph
npm ci
npm run check
npm startnpm start serves stdio MCP. npm run check runs the software tests, build, all-tool example calls, semantic audit, dependency audit, and package dry run. For a host, point command at node and args at the absolute path to dist/index.js.
How it works
flowchart LR
A["FASTA · PDB · SMILES"] --> H["MCP host"]
H --> S["ImmunoGraph<br/>46 typed tools"]
S --> D["Local calculations<br/>and exact fixtures"]
S --> L["Optional live<br/>connectors"]
D --> R["Results + provenance"]
L --> R
R --> O["Reports · 3D files · ZIP"]The host chooses which tools to call. ImmunoGraph validates typed inputs, executes a handler, and returns structured success or failure data with run metadata and source labels. The server does not silently run a full study or treat fixture values as live evidence. See architecture for transport and module details.
Technical design
Concern | Implementation | Why it matters |
Client interoperability | Official MCP SDK; stdio for local hosts and guarded Streamable HTTP for deployments | One discoverable tool contract works across MCP-capable clients |
Input and result contracts | Zod schemas, typed handlers, structured MCP results, explicit errors | Clients can validate arguments and handle failures programmatically |
Scientific provenance | Source status, method/run metadata, and versioned exact-match fixtures | Reviewers can distinguish configured live work from offline demonstrations |
Export integrity | Inline artifact bytes, byte lengths, and SHA-256 digests | A host can save actual PDB/PDBQT/CSV/ZIP files and verify them |
Operational boundaries | Optional connectors and local scientific binaries are configured separately | Missing providers fail explicitly or use only the requested, labelled fallback |
The registry entry points to the published OCI image. Optional scientific binaries, model downloads, and third-party service availability remain separate runtime requirements.
A research workflow
flowchart LR
FASTA["Protein FASTA"] --> V["Peptide windows"]
V --> E["Immune evidence"]
E --> R["Constraints + ranking"]
R --> X["Reviewable exports"]
E -.->|Optional| S["Structure + docking"]
S -.-> XThese are separate MCP calls. Rankings and reports depend on the evidence the client supplies. With live prediction disabled, an arbitrary new protein has no automatic scientific binding result.
Docking and 3D files
Configured live tools can use actual coordinates rather than decorative models:
flowchart LR
P["Protein PDB"] --> R["Meeko receptor<br/>PDBQT"]
L["Compound / SMILES"] --> M["RDKit + Meeko ligand<br/>SDF + PDBQT"]
R --> V["AutoDock Vina"]
M --> V
V --> POSE["Pose PDBQT<br/>complex PDB"]
POSE --> VIEW["3D viewer HTML"]
POSE --> PLIP["PLIP contacts"]
POSE --> ZIP["Research ZIP"]Real 3D outcomes
1IEP + imatinib: exported top Vina pose. This figure is rendered from a saved complex PDB produced by a LIVE_ONLY MCP docking call. Panel A traces the protein Cα coordinates and shows the ligand; panel B shows the ligand and protein atoms within 5 Å. The render script reproduces the image from those coordinates. This is a docking capability example, not an epitope candidate; the pose score is a computational estimate, not measured affinity.

1UYD + PubChem CID 2244: three computed poses. This figure comes from the live docking research ZIP, which includes the receptor, ligand, Vina poses, complex, and image. Its run provenance and summary identify the sources and nine generated poses; the image shows three of them. The callouts show distances calculated from the coordinates; dashed polar contacts are inferred geometry, not experimentally established bonds.
The live docking guide documents inputs, runtime requirements, example files, and verification of the exported pose chain. An interactive 3D viewer visualizes returned coordinate data; neither image establishes experimental binding.
Provenance you can inspect
Result path | Meaning | Scientific use |
| A configured external provider or local scientific runtime executed on supplied inputs | Review its method, parameters, and limitations |
| An approved, exact-match offline replay | Demonstration only |
| Deterministic teaching data generated by a demo tool | Demonstration only |
Caller-supplied calculations | Ranking, constraints, calibration, and reporting transform supplied evidence | Depends on the quality and provenance of that evidence |
The 46-tool reality matrix records what each tool actually computes or fetches and the verification performed. In particular, GraphBepi is fixture-only in this release; Parker is a hydrophilicity calculation; BepiPred-3.0 is the supported live B-cell option when enabled.
Tool catalog
Group | Tools | Examples |
Prediction | 6 | FASTA validation, peptide windows, MHC-I/II, B-cell |
Evidence and prioritization | 9 | Score normalization, consensus, HLA coverage, ranking |
Constraints | 5 | Overlap, duplicates, thresholds, candidate rules |
Structure | 7 | RCSB/AlphaFold, accessibility, pockets, Mol* |
Chemistry | 5 | PubChem, RDKit descriptors, ligand preparation |
Docking | 5 | Receptor preparation, Vina, pose clustering, PLIP |
Reports and utilities | 9 | CSV/report data, explanations, workflow trace, ZIP |
Total | 46 | One MCP server |
See the complete tool catalog and input contracts. To try one call, invoke validate_sequence with:
{
"fasta": ">protein\nACDEFGHIK",
"profileVersion": "mvp-v1.0"
}It returns the normalized sequence, length, and SHA-256 identity, or a structured validation error.
Enable live connectors
Most live connectors are disabled by default. Set flags in the MCP server's environment and install the relevant provider runtime or allow its network endpoint. .env.example lists the settings.
Capability | Switch | Additional requirement |
IEDB MHC-I/II, BepiPred-3.0, population coverage |
| Access to the relevant IEDB APIs |
RCSB PDB, AlphaFold DB, PubChem |
| Provider network access |
RDKit, Meeko, FreeSASA, fpocket | Installed Python packages or native binaries | |
AutoDock Vina and PLIP |
| Installed runtimes and valid coordinate inputs |
If a required provider or binary is unavailable, the tool returns an explicit failure or a labelled fallback according to the requested policy. The npm tarball and container do not bundle optional scientific binaries or model downloads.
HTTP deployment
npm run start:http serves http://127.0.0.1:3000/mcp locally, with GET /health as a health probe. For a remote endpoint, set HOST=0.0.0.0, a long IMMUNOGRAPH_API_KEY, exact IMMUNOGRAPH_ALLOWED_HOSTS and IMMUNOGRAPH_ALLOWED_ORIGINS, and terminate TLS at your ingress. The image runs the HTTP entry point by default; publishing the image does not create a managed endpoint. Read SECURITY.md before deployment.
Verification and scope
The CI workflow checks Windows and Linux on Node.js 20 and 22. A standard MCP client has exercised all 46 documented examples; the semantic audit checks calculations, provenance, and artifact hashes. The v0.2.1 container release run tested 46-tool discovery and a real call over Docker stdio, guarded HTTP, and a test TLS ingress. The protocol audit also checks that every tool advertises parameter descriptions and a structured result schema. These checks establish software behavior and packaging, not the accuracy of every scientific method or continuous availability of third-party services.
The motivation is familiar in epitope research: studies such as Grifoni et al. (2020) and Ferreira et al. (2021) combine sequence analysis, prediction, conservation, and other evidence across tools. ImmunoGraph makes computational handoffs easier to inspect; independent review and experimental validation remain necessary.
Contribute and get help
Contributions and reproducible bug reports are welcome. Start with CONTRIBUTING.md, open an issue, and report security concerns through SECURITY.md. Released changes are in CHANGELOG.md. ImmunoGraph is available under the MIT license.
Available Tools
46 toolsapply_constraint_rulesARead-onlyIdempotent
Apply base, duplicate, and overlap rules to one immutable candidate snapshot and return final rule outcomes. Use after validate_thresholds, remove_duplicate_candidates, and detect_overlapping_epitopes when their inputs are ready. The snapshot hash binds the decision to supplied evidence; this tool does not fetch new evidence.
Inputs: runId (required), snapshotHash (required), baseConstraints (required), overlapThreshold (required), +3 more. Returns: ok, data, meta, error
calculate_molecular_descriptorsARead-onlyIdempotent
Calculate molecular descriptor values for a supplied SMILES using the selected RDKit or lightweight method. Use after validate_compound for chemistry features; this does not predict binding or prepare 3D ligand coordinates. RDKit mode requires its runtime, while lightweight estimates are explicitly identified as such.
Inputs: runId (required), method (required), smiles (required), compoundId (required). Returns: ok, data, meta, error
calculate_population_coverageARead-onlyIdempotent
Estimate HLA population coverage for supplied peptide-allele associations using a configured authoritative IEDB service or an explicitly allowed exact fixture. Use when the allele restrictions and population IDs are known; use calculate_synthetic_population_coverage only for demonstrations. LIVE_ONLY requires the live service and the result carries source provenance.
Inputs: associations (required), populationIds (required), runId, classMode, +1 more. Returns: ok, data, meta, error
calculate_structure_confidenceARead-onlyIdempotent
Extract per-residue AlphaFold pLDDT from supplied AlphaFold PDB bytes or summarize supplied confidence metrics. Use to inspect local model confidence, not solvent exposure or epitope probability; experimental PDB B-factors are not accepted as pLDDT. Returns residue scores or a metric summary with source context.
Inputs: runId (required), structureId (required), sourceStatus (required), modelSource, +3 more. Returns: ok, data, meta, error
calculate_surface_accessibilityARead-onlyIdempotent
Calculate relative solvent accessibility for mapped residues with FreeSASA on supplied PDB coordinates, or replay an exact fixture. Use after map_epitopes_to_structure when asking whether residues are exposed; use calculate_structure_confidence for model pLDDT instead. Live computation requires FreeSASA and PDB bytes; output is computed accessibility, not epitope probability.
Inputs: runId (required), method (required), mappings (required), structurePdbBase64. Returns: ok, data, meta, error
calculate_synthetic_population_coverageARead-onlyIdempotent
Calculate deterministic demonstration coverage from synthetic HLA-frequency inputs. Use for teaching or testing the coverage pipeline, never as a population estimate; use calculate_population_coverage for a configured authoritative source. The output remains synthetic regardless of its numeric values.
Inputs: associations (required), populationIds (required), runId, classMode. Returns: ok, data, meta, error
calibrate_confidenceARead-onlyIdempotent
Fit Platt or isotonic calibration on labeled samples and evaluate on separate labeled outcomes, or emit rule-based confidence. Use fitted modes only with real calibration and evaluation labels; rule-based mode has no measured reliability. Returns calibrated probabilities and metrics only where the input labels support them.
Inputs: runId (required), method (required), predictions (required), evaluationSamples, +1 more. Returns: ok, data, meta, error
categorize_candidatesARead-onlyIdempotent
Assign provisional decision categories and blocking conditions from supplied scores, rules, and thresholds. Use after scoring and rule evaluation; use rank_candidates for numeric ordering, not category assignment. Returns deterministic categories that retain the input evidence boundaries.
Inputs: runId (required), candidates (required), thresholds (required). Returns: ok, data, meta, error
chat_with_research_agentARead-onlyIdempotent
Answer a researcher question from the supplied evidence summary under the bounded agent policy. Use for an explanation after evidence has been assembled; this tool does not independently retrieve papers, predict epitopes, or run the workflow. Deterministic mode cites supplied evidence; LLM mode requires configured model access.
Inputs: runId (required), question (required), agentMode (required), evidenceSummary (required). Returns: ok, data, meta, error
cluster_docking_posesARead-onlyIdempotent
Group supplied Vina ligand poses by coordinate RMSD using DBSCAN or hierarchical clustering, or replay an exact fixture. Use after run_docking to summarize pose diversity; this does not redock or measure affinity. Direct mode requires matching pose records and PDBQT bytes and returns cluster assignments.
Inputs: poses (required), runId (required), method (required), dockingRunId (required), +2 more. Returns: ok, data, meta, error
compute_consensusARead-onlyIdempotent
Combine one candidate group of normalized observations into weighted consensus, agreement, and completeness. Use after normalize_scores when the group and required weight are known; use compute_consensus_batch for several independent groups. It calculates from caller-supplied evidence and does not make a new prediction.
Inputs: runId (required), groupKey (required), observations (required), configuredRequiredWeight (required). Returns: ok, data, meta, error
compute_consensus_batchARead-onlyIdempotent
Compute weighted consensus separately for multiple candidate groups in one call. Use instead of repeated compute_consensus calls when groups are already assembled; groups are not pooled together. Returns one result per supplied group, based only on its input observations.
Inputs: runId (required), groups (required). Returns: ok, data, meta, error
create_molstar_viewARead-onlyIdempotent
Package supplied protein and optional ligand PDB coordinates into an interactive Mol* HTML viewer, or return an exact fixture reference. Use when a client needs a 3D visualization after structure or docking work; this does not calculate a new pose. Returns HTML bytes whose viewer loads its library from a CDN.
Inputs: mode (required), runId (required), viewId (required), structureArtifactRef (required), +3 more. Returns: ok, data, meta, error
deduplicate_compoundsARead-onlyIdempotent
Compare supplied compounds by RDKit canonical isomeric SMILES and InChIKey when live chemistry is requested. Use after validate_compound to collapse chemical identity duplicates; this is distinct from remove_duplicate_candidates, which handles epitope positions. Returns canonical groups and preserves input provenance.
Inputs: runId (required), compounds (required). Returns: ok, data, meta, error
describe_agentic_workflowARead-onlyIdempotent
Describe the bounded workflow manifest, stage permissions, and approval gates without running tools. Use to inspect orchestration before run_agentic_workflow; this is a description, not a live scientific call. Returns the declared stages and tool permissions.
Inputs: runId (required), runIntent (required), includeFutureInterfaces (required). Returns: ok, data, meta, error
detect_binding_pocketsARead-onlyIdempotent
Run fpocket on supplied PDB coordinates to identify geometric binding pockets, or replay an exact fixture. Use after structure validation for pocket exploration; use run_docking to score a prepared receptor-ligand pair. Live mode needs the fpocket binary and PDB bytes; pocket scores are computational descriptors.
Inputs: runId (required), method (required), structureId (required), structureArtifactRef (required), +1 more. Returns: ok, data, meta, error
detect_overlapping_epitopesARead-onlyIdempotent
Find positional overlap competitors and connected components among candidate intervals. Use before apply_constraint_rules when overlaps need review; this tool does not choose a winner or deduplicate identities. Returns interval relationships from caller-supplied coordinates.
Inputs: runId (required), threshold (required), candidates (required). Returns: ok, data, meta, error
explain_candidateARead-onlyIdempotent
Explain a fixed candidate category and score from supplied component scores, rule outcomes, and provenance. Use after ranking and constraints when a researcher needs a decision rationale; this cannot change scientific values. Deterministic mode uses only supplied evidence; optional LLM paraphrase needs its configured runtime.
Inputs: runId (required), audience (required), category (required), trackRank (required), +6 more. Returns: ok, data, meta, error
export_candidatesARead-onlyIdempotent
Serialize supplied candidate records into JSON or CSV with raw and normalized scores kept separate. Use after candidate review for a table export; generate_report produces a broader report. Returns file bytes and checksum metadata, not new scientific scores.
Inputs: runId (required), format (required), categories (required), idempotencyKey (required), +1 more. Returns: ok, data, meta, error
export_research_packageARead-onlyIdempotent
Build a checksummed research ZIP from a supplied package snapshot and selected sections. Use after reports and artifacts are ready for handoff; generate_report exports individual report files. It serializes supplied evidence and metadata rather than launching predictors or docking.
Inputs: runId (required), idempotencyKey (required), includeDocking (required), includeChemistry (required), +3 more. Returns: ok, data, meta, error
export_workflow_traceARead-onlyIdempotent
Serialize supplied ordered workflow events into a redacted JSON trace artifact. Use for audit or reproducibility after a workflow; generate_report is for the scientific summary. Returns inline bytes and SHA-256 metadata while omitting secrets and private reasoning.
Inputs: runId (required), idempotencyKey (required), redactionProfile (required), traceEvents. Returns: ok, data, meta, error
extract_interactionsARead-onlyIdempotent
Run PLIP on a supplied protein-ligand complex PDB to identify computed contact geometry, or replay an exact fixture. Use after run_docking when a representative pose has been exported; this does not experimentally measure binding. Live mode requires PLIP and returns structured contacts plus the original XML artifact.
Inputs: runId (required), method (required), dockingRunId (required), representativePoseIds (required), +1 more. Returns: ok, data, meta, error
fetch_compoundARead-onlyIdempotent
Retrieve compound identity and SMILES from configured PubChem access or replay an exact fixture. Use with a known compound reference before validation or ligand preparation; this does not calculate descriptors. LIVE_ONLY requires PubChem access and returns source provenance.
Inputs: runId (required), source (required), compoundRef (required), fallbackPolicy (required). Returns: ok, data, meta, error
fetch_structureARead-onlyIdempotent
Retrieve PDB coordinates from RCSB or an AlphaFold model from the configured provider, or replay an exact fixture. Use when a structure accession is known and before coordinate-based analyses; this is not sequence prediction. LIVE_ONLY requires network access and returns source provenance plus coordinate artifact bytes.
Inputs: runId (required), source (required), targetId (required), accession (required), +1 more. Returns: ok, data, meta, error
generate_candidate_peptidesARead-onlyIdempotent
Enumerate one-based peptide windows from a supplied protein sequence for MHC-I or MHC-II exploration. Use after validate_sequence and before prediction; this creates candidates, not binding scores. Returns each window with peptide, start, end, and length.
Inputs: sequence (required), candidateType (required), peptideLengths (required), runId, +1 more. Returns: ok, data, meta, error
generate_reportARead-onlyIdempotent
Render the supplied approved report snapshot into requested JSON, Markdown, or CSV artifacts. Use after evidence review when a report is needed; export_candidates is for candidate rows and export_research_package is for a ZIP. Returns inline file bytes with lengths and SHA-256 hashes; it does not independently fetch evidence.
Inputs: runId (required), outputFormats (required), idempotencyKey (required), templateVersion (required), +1 more. Returns: ok, data, meta, error
map_epitopes_to_structureARead-onlyIdempotent
Align supplied candidate sequence positions to residue coordinates in a PDB structure, or replay an exact fixture. Use after validate_structure when mapping sequence evidence onto a model; use calculate_surface_accessibility for solvent exposure. Direct-coordinate mode requires reference sequence and PDB bytes and returns mapped positions.
Inputs: runId (required), candidates (required), mappingMode (required), structureId (required), +2 more. Returns: ok, data, meta, error
normalize_scoresARead-onlyIdempotent
Transform supplied raw observations into comparable unit-interval values using a named versioned normalization registry. Use after evidence collection and before consensus or ranking; this does not fetch predictors or improve source quality. Returns normalized observations and records the applied transformation.
Inputs: runId (required), observations (required), registryVersion (required). Returns: ok, data, meta, error
optimize_construct_geneticARead-onlyIdempotent
Run seeded genetic-style search over supplied candidates to balance construct size, coverage, redundancy, and manufacturability terms. Use after final ranking when searching construct arrangements; use optimize_shortlist_coverage for simpler deterministic shortlist selection. Returns a computed proposal, not a validated vaccine construct.
Inputs: runId (required), linker (required), candidates (required), populationIds (required), +6 more. Returns: ok, data, meta, error
optimize_shortlist_coverageARead-onlyIdempotent
Select a deterministic T-cell shortlist from eligible ranked candidates using supplied population coverage and construct constraints. Use after final ranking when shortlist size or coverage is the decision; use optimize_construct_genetic for seeded construct search. The result depends on the supplied coverage matrix and is not an independently measured population outcome.
Inputs: runId (required), method (required), populationIds (required), eligibleCandidateIds (required), +6 more. Returns: ok, data, meta, error
predict_bcellARead-onlyIdempotent
Score linear B-cell sequence positions using configured live IEDB BepiPred-3.0, local Parker hydrophilicity, or an exact GraphBepi fixture. Choose the method explicitly: Parker is a physicochemical proxy and GraphBepi is fixture-only in this release. Returns residue scores, regions, and provenance; live IEDB access may fail or time out.
Inputs: runId (required), methods (required), parameters (required), proteinRef (required), +2 more. Returns: ok, data, meta, error
predict_mhciARead-onlyIdempotent
Get MHC-I peptide-binding observations for supplied class-I alleles and peptide lengths through the configured IEDB connector or an allowed exact fixture. Use for class-I evidence, not MHC-II or B-cell evidence. LIVE_ONLY requires the live connector; results include method and source provenance, and unavailable sources return a structured failure.
Inputs: alleles (required), peptideLengths (required), runId, methods, +3 more. Returns: ok, data, meta, error
predict_mhciiARead-onlyIdempotent
Get MHC-II peptide-binding observations for supplied class-II alleles and peptide lengths through the configured IEDB connector or an allowed exact fixture. Use for class-II evidence rather than predict_mhci or predict_bcell. LIVE_ONLY requires the live connector; results include method and source provenance, and unavailable sources return a structured failure.
Inputs: alleles (required), peptideLengths (required), runId, methods, +3 more. Returns: ok, data, meta, error
predict_synthetic_bindingARead-onlyIdempotent
Create deterministic synthetic MHC binding values for demos and software tests from supplied candidates and alleles. Use only when teaching or testing the pipeline; use predict_mhci or predict_mhcii for configured scientific prediction. Output is marked synthetic and must not be treated as binding evidence.
Inputs: runId (required), method (required), alleles (required), candidates (required), +4 more. Returns: ok, data, meta, error
prepare_ligandARead-onlyIdempotent
Generate a 3D SDF and Meeko PDBQT from supplied SMILES using configured RDKit and Meeko, or return an exact fixture reference. Use before run_docking for a ligand; calculate_molecular_descriptors does not produce docking coordinates. Live preparation returns artifact bytes and hashes and fails if a required runtime is absent.
Inputs: runId (required), smiles (required), compoundId (required), preparationMethod (required). Returns: ok, data, meta, error
prepare_receptorARead-onlyIdempotent
Convert supplied protein PDB coordinates and selected chains into receptor PDBQT with Meeko, or return an exact fixture reference. Use before run_docking for the receptor; validate_structure checks coordinates but does not prepare charges or PDBQT. Live preparation requires the Meeko runtime and returns artifact bytes.
Inputs: runId (required), chainIds (required), structureId (required), preparationMethod (required), +1 more. Returns: ok, data, meta, error
rank_candidatesARead-onlyIdempotent
Calculate preliminary candidate scores or final stable ranks from supplied evidence snapshots and completion flags. Use PRELIMINARY before final constraints and FINAL only after base and final constraints are complete; this does not call predictors. Returns ranked candidates and applies the versioned ranking profile to the supplied values.
Inputs: phase (required), runId (required), candidates (required), rankingProfileVersion (required), +3 more. Returns: ok, data, meta, error
remove_duplicate_candidatesARead-onlyIdempotent
Group candidates with identical positional identity and choose canonical records. Use before apply_constraint_rules to remove true duplicates; identical peptide strings at different coordinates remain separate. Returns canonical and duplicate relationships without changing source evidence.
Inputs: runId (required), candidates (required), proteinHash (required). Returns: ok, data, meta, error
run_agentic_workflowARead-onlyIdempotent
Execute caller-supplied approved typed toolCalls in bounded stage order, or return a plan if no calls are supplied. Use when the client has already selected inputs and allowed tools; describe_agentic_workflow only inspects the manifest. The tool does not autonomously invent calls, and approval gates pause rather than self-approve.
Inputs: runId (required), agentMode (required), objective (required), approvedToolNames (required), +2 more. Returns: ok, data, meta, error
run_dockingARead-onlyIdempotent
Run configured AutoDock Vina on receptor and ligand PDBQT bytes inside a supplied box, or replay an exact fixture. Use after prepare_receptor, prepare_ligand, and validate_docking_box; LIVE_ONLY requires Vina and never silently substitutes a fixture. Returns scored coordinate poses, viewer HTML, and ZIP bytes; runtime can be substantial and scores are computational estimates.
Inputs: mode (required), runId (required), ligandId (required), receptorId (required), +9 more. Returns: ok, data, meta, error
validate_compoundARead-onlyIdempotent
Parse and canonicalize supplied SMILES with RDKit for live-sourced compounds, or validate an explicit fixture record. Use before deduplication, descriptors, or ligand preparation; this checks chemical syntax and identity, not biological activity. RDKit unavailability returns a structured failure for the live path.
Inputs: runId (required), smiles (required), compoundId (required), sourceStatus (required). Returns: ok, data, meta, error
validate_docking_boxARead-onlyIdempotent
Validate the center and size of a proposed docking search box and calculate its volume. Use before run_docking to catch invalid box geometry; this does not run Vina or establish a binding site. Returns validated dimensions or a structured validation failure.
Inputs: size (required), runId (required), center (required), dockingBoxId (required). Returns: ok, data, meta, error
validate_sequenceARead-onlyIdempotent
Validate one protein FASTA record and return its normalized sequence, length, header, and SHA-256 identity. Use before peptide generation or prediction; this checks sequence syntax, not structure or biological activity. Invalid input returns a structured validation error.
Inputs: fasta (required), profileVersion (required). Returns: ok, data, meta, error
validate_structureARead-onlyIdempotent
Check supplied structure metadata and, when PDB bytes are provided, parse actual atoms, chains, and residues. Use before mapping, accessibility, or docking preparation; this does not evaluate model confidence or binding. Mismatched metadata returns a structured failure.
Inputs: runId (required), format (required), chainIds (required), structureId (required), +3 more. Returns: ok, data, meta, error
validate_thresholdsARead-onlyIdempotent
Evaluate the versioned hard biological threshold profile against supplied candidates. Use before categorization or final constraint application; this reports rule outcomes rather than ranking candidates. Missing evidence is represented in the rule results instead of invented as a pass.
Inputs: runId (required), candidates (required), ruleProfileVersion (required). Returns: ok, data, meta, error
visualize_resultsARead-onlyIdempotent
Build a deterministic visualization view model from supplied evidence for the selected chart or graph type. Use when a UI will render results; create_molstar_view is for coordinate-based 3D HTML. Returns view-model data, not a PNG or an independent scientific analysis.
Inputs: runId (required), visualizationType (required), visualizationSnapshot. Returns: ok, data, meta, error
Tool Schema Changelog
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
TDQS
Scored across 46 tools
Despite 46 tools, descriptions consistently state scope boundaries and name explicit alternatives (predict_mhci vs predict_mhcii vs predict_bcell, compute_consensus vs compute_consensus_batch, remove_duplicate_candidates vs deduplicate_compounds, generate_report vs export_candidates). A few near-neighbors remain easy to confuse at a glance—the synthetic vs authoritative pairs, and create_molstar_view vs visualize_results vs generate_report—but the prose resolves them. No two tools appear to do the same thing.
Every tool is snake_case with a verb-first, object-second pattern (validate_sequence, calculate_surface_accessibility, prepare_ligand, run_docking, export_workflow_trace). Suffix modifiers like _batch and prefixes like compute_/calculate_ are used consistently within their families. Minor quirks (predict_mhci without a separator before the numeral) do not break predictability.
The 46 tools map onto genuinely distinct pipeline stages—sequence, prediction, consensus, constraints, structure, chemistry, docking, reporting, and orchestration—so most earn their place rather than being padding. Still, 46 is heavy for a single MCP surface and forces an agent to plan across many stages, with several parallel 'synthetic/demo' twins adding surface area.
The surface covers the full lifecycle from sequence validation through prediction, normalization, consensus, ranking, constraints, structure mapping, docking, interaction analysis, and reporting/orchestration export. Gaps are minor: no allele/population listing or lookup tools, no async job status or artifact retrieval, and no cleanup or cache management operations.
Maintenance
Related MCP Connectors
Create, browse, remix, collaborate on, and run durable AI workflow nodes from MCP hosts.
Workflow diagnostics, capability routing, and x402 settlement for MCP-compatible agents.
Governed MCP: agent audit, provenance, deterministic checks, and receipt-backed FragGate execution.
Public MCP endpoint for collaborative research and knowledge discovery.
Related MCP Servers
- AlicenseNot gradedqualityCmaintenanceMCP-native scientific skills for reproducible computational biology and AI-driven drug-discovery workflows. It combines deterministic scientific tools with an MCP server to give AI agents real computational capabilities.Apache 2.0
- AlicenseAqualityCmaintenanceEnables AI assistants to run reproducible bioinformatics pipelines over MCP, with verifiable provenance via checksums and Workflow Run RO-Crate metadata.62MIT
- AlicenseAqualityCmaintenanceProvides MCP access to plan and execute reproducible computational experiments with immutable execution plans, artifact provenance, verification checks, and support for direct, Slurm, and PBS backends.17Apache 2.0
- AlicenseBqualityCmaintenanceEnables biomedical research workflows via MCP, including experiment and document search, PDF text extraction, evidence retrieval, and validation with provenance tracking.10MIT