MCP Tools Reference

The Hubify MCP server exposes 48 tools across two categories:

Category	Count	Prefix	Use for
Platform tools	31	none	Reading and writing lab data directly
Phase E ledger tools	17	`hubify_`	Writing to the agentEvents ledger; surfaces in Captain View

All tools work without authentication for both read and write operations when the MCP server is configured with CONVEX_URL and HUBIFY_LAB_SLUG.

Platform Tools

`lab_create`

Create a new lab and automatically provision a matching GitHub repository. If the user has a connected GitHub account, the repo is created in their own account. Otherwise it is created in the Hubify-Projects org.

Parameter	Type	Required	Description
`userId`	string	Yes	Convex user ID
`name`	string	Yes	Lab name (used as repo name base)
`description`	string	No	Short description
`mission`	string	No	Mission statement
`subdomain`	string	No	URL subdomain (auto-derived from name if omitted)
`isPublic`	boolean	No	Whether the lab and repo are public (default: true)

Returns: { labId, repoUrl } — repoUrl is null if GitHub repo creation failed (lab is still created).

`lab_status`

Get a snapshot of the current lab: experiment count, paper count, agent count, pod count, anomaly count. No parameters.

`lab_list`

List every lab on this Convex deployment with slug, name, status, and which is currently active. No parameters.

`lab_switch`

Hot-swap the active lab for the current MCP session without restarting.

Parameter	Type	Required	Description
`slug`	string	No	Lab slug to switch to. Omit to reset to the env-configured default.

Experiments

`experiment_list`

List experiments with status, title, and runtime.

Parameter	Type	Required	Description
`status`	string	No	Filter: `queued`, `running`, `pass`, `fail`, `blocked`

`experiment_create`

Create and queue a new experiment (auto-generates EXP-NNN ID).

Parameter	Type	Required	Description
`title`	string	Yes	Experiment title
`config`	string	No	JSON config string (hyperparameters, etc.)
`template`	string	No	Template name to base the experiment on
`computeMode`	string	No	`gpu`, `cpu`, or `distributed`. Default: `gpu`

`experiment_run`

Dispatch a queued experiment to a GPU pod. Accepts EXP-* ID or Convex document _id.

Parameter	Type	Required	Description
`experimentId`	string	Yes	`EXP-BB-001` or Convex `_id`
`podId`	string	No	Target pod ID. Auto-picks a running pod if omitted.

`experiment_queue`

List queued and running experiments. No parameters.

Papers

`paper_list`

List all papers with readiness percentages and status. No parameters.

`paper_create`

Create a new paper draft.

Parameter	Type	Required	Description
`title`	string	Yes	Paper title
`abstract`	string	No	Paper abstract
`target`	string	No	Target journal (e.g. `Physical Review D`, `MNRAS`)
`status`	string	No	`draft`, `review`, `submitted`, `published`. Default: `draft`

`paper_review`

Trigger a cross-model peer review on a paper via the orchestrator.

Parameter	Type	Required	Description
`paperName`	string	Yes	Paper name or title

Tasks

`task_list`

List tasks with status and priority.

Parameter	Type	Required	Description
`status`	string	No	Filter: `backlog`, `todo`, `in_progress`, `review`, `done`

`task_create`

Create a new task.

Parameter	Type	Required	Description
`title`	string	Yes	Task title
`description`	string	No	Task details
`priority`	string	No	`low`, `medium`, `high`, `urgent`. Default: `medium`
`status`	string	No	Initial status. Default: `todo`

`task_update`

Update task status or priority.

Parameter	Type	Required	Description
`taskId`	string	Yes	Convex task `_id`
`status`	string	No	New status
`priority`	string	No	New priority

Memory & Knowledge

`memory_search`

Search across all 4 memory layers: user messages, agent learnings, knowledge wiki, and lab decisions.

Parameter	Type	Required	Description
`query`	string	Yes	Search query
`layer`	string	No	`user`, `agent`, `lab`, `decisions`, or `all`. Default: `all`

`memory_save`

Save to the lab memory system.

Parameter	Type	Required	Description
`type`	string	Yes	`preference`, `decision`, or `message`
`key`	string	Conditional	Required for `preference` type
`value`	string	Conditional	Required for `preference` type
`title`	string	Conditional	Required for `decision` type
`decision`	string	Conditional	Required for `decision` type
`rationale`	string	Conditional	Required for `decision` type
`content`	string	Conditional	Required for `message` type
`summary`	string	No	One-line summary for `message` type

`knowledge_search`

Full-text search across the lab’s knowledge base: experiments, papers, agents, surveys, knowledge entries.

Parameter	Type	Required	Description
`query`	string	Yes	Search query

`knowledge_add`

Add an entry to the lab’s knowledge wiki.

Parameter	Type	Required	Description
`name`	string	Yes	Entry name (e.g. `Hellings-Downs Correlation`)
`entityType`	string	Yes	`concept`, `entity`, `source`, `method`, `result`, or `dataset`
`description`	string	No	Entry description
`relatedEntities`	string[]	No	Related entity names

`agent_learning_save`

Save an operational or scientific learning to agent memory.

Parameter	Type	Required	Description
`key`	string	Yes	Short unique key (e.g. `mcmc-convergence-criteria`)
`insight`	string	Yes	What was learned — specific and actionable
`type`	string	No	`operational`, `scientific`, `procedural`. Default: `operational`
`confidence`	number	No	Confidence 0–1. Default: 0.7

Contributions

`contribution_list`

List scientific contributions with novelty scores.

Parameter	Type	Required	Description
`limit`	number	No	Max results. Default: 20

`contribution_create`

Log a scientific contribution.

Parameter	Type	Required	Description
`title`	string	Yes	Contribution title
`description`	string	Yes	What was discovered or contributed
`type`	string	No	`prediction`, `measurement`, `method`, `catalog`, `constraint`, `null_result`. Default: `measurement`

Agents & Communication

`agent_list`

List all agents with roles, status, and models. No parameters.

`chat`

Send a message to the lab orchestrator and wait for its response. Shares history with the web chat and CLI surfaces.

Parameter	Type	Required	Description
`message`	string	Yes	Message to send
`sessionId`	string	No	Existing session ID to resume. Pass `"new"` to start fresh.
`waitMs`	number	No	Max wait for response in ms. Default: 90000

Compute & Costs

`pod_list`

List GPU pods with status, GPU type, and cost. No parameters.

`pod_create`

Create a new GPU pod via RunPod API.

Parameter	Type	Required	Description
`gpuType`	string	No	GPU type (e.g. `H200`, `H100`, `A100`). Default: `H100`
`name`	string	No	Pod name

`cost_summary`

Get daily/weekly/monthly/all-time cost breakdown vs budget. No parameters.

Misc

`notification_list`

List recent notifications (alerts, experiment completions, agent messages).

Parameter	Type	Required	Description
`limit`	number	No	Max results. Default: 20
`unreadOnly`	boolean	No	Only unread notifications. Default: false

`survey_list`

List astronomical surveys with anomaly counts and source statistics. No parameters.

`figure_list`

List all figures with titles, types, and image URLs.

Parameter	Type	Required	Description
`limit`	number	No	Max results. Default: 20

`standup_trigger`

Trigger an agent standup.

Parameter	Type	Required	Description
`type`	string	No	`morning`, `midday`, `evening`. Default: `morning`

`activity_list`

List recent lab activity (experiment completions, agent actions, alerts).

Parameter	Type	Required	Description
`limit`	number	No	Max results. Default: 20

Phase E Ledger Tools (`hubify_*`)

These tools write to the agentEvents ledger. Events appear in real time in the Hubify web app’s Captain View and activity feed. Use these when Claude Code is acting as a local lab operator.

Context & Session

`hubify_get_context`

Full lab snapshot as structured JSON: lab metadata, stats, recent events, contributions, papers. Use at session start.

Parameter	Type	Required	Description
`recentLimit`	number	No	Recent events to include. Default: 25

`hubify_get_context_pack`

Token-budgeted plain-text context pack for agent system prompts (lab state + in-flight experiments + recent events).

Parameter	Type	Required	Description
`target`	string	No	Compaction style: `claude`, `codex`, `generic`. Default: `generic`
`maxBytes`	number	No	Approximate output cap. Default: 8000

`hubify_create_session_summary`

Persist a session summary to the ledger. Use at the end of every Claude Code session.

Parameter	Type	Required	Description
`summary`	string	Yes	Markdown or plain-text summary of what was done
`sessionId`	string	No	Session identifier. Auto-generates a UUID if omitted.
`durationMs`	number	No	Session duration in ms
`surface`	string	No	Where the session ran: `claude-code`, `cli`, `sidebar`

Event Logging

`hubify_log_event`

Append a typed event to the agentEvents ledger. Idempotent on clientEventId. Common event types: paper.section.drafted, experiment.completed, discovery.proposed, run.recorded, agent.woke, agent.sleeping.

Parameter	Type	Required	Description
`type`	string	Yes	Dotted event type (e.g. `paper.section.drafted`)
`clientEventId`	string	No	Idempotency key. Auto-generated if omitted.
`actorKind`	string	No	`captain`, `agent`, `orchestrator`, `system`, `webhook`, `cli`, `mcp`. Default: `mcp`
`actorLabel`	string	No	Human-readable actor name (e.g. `claude-code`)
`entityType`	string	No	Entity the event is about (`paper`, `experiment`, `claim`)
`entityId`	string	No	Convex `_id` or external ID of the entity
`payload`	object	No	JSON-encodable body
`severity`	string	No	`debug`, `info`, `warn`, `error`. Default: `info`
`correlationId`	string	No	Links related events together
`parentEventId`	string	No	Parent event for nested flows

`hubify_log_finding`

Log a scientific finding. When noveltyScore >= 6, also creates a contributions row visible in the Contributions UI.

Parameter	Type	Required	Description
`summary`	string	Yes	One-paragraph summary of the finding
`noveltyScore`	number	No	1–10 novelty rating. Default: 5
`claimIds`	string[]	No	Related claim IDs
`experimentId`	string	No	Source experiment Convex `_id`
`type`	string	No	`method`, `dataset`, `result`, `theorem`. Default: `result`

`hubify_sync_events`

Batch-flush locally-buffered events in one round-trip. Idempotent on clientEventId.

Parameter	Type	Required	Description
`events`	object[]	Yes	Array of event objects, each with `type` and `clientEventId` required

Each event in the array accepts: type, clientEventId, actorKind, actorLabel, entityType, entityId, payload, severity, correlationId, occurredAt.

Run & Claims

`hubify_record_run`

Record a new run directory, tying paths to the originating experiment. Emits run.recorded.

Parameter	Type	Required	Description
`runId`	string	Yes	Run ID, e.g. `2026-04-29-mcmc-final`
`paths`	string[]	Yes	Files written (relative to repo root)
`experimentId`	string	No	Source experiment `_id`
`summary`	string	No	Brief summary of what the run produced

`hubify_create_claim`

Create a scientific claim. Emits claim.proposed and creates a contributions row.

Parameter	Type	Required	Description
`text`	string	Yes	The claim — one falsifiable assertion
`paperId`	string	No	Paper this claim belongs to
`experimentId`	string	No	Source experiment `_id`
`noveltyScore`	number	No	1–10. Default: 5

`hubify_update_claim`

Update a claim’s status. Emits claim.upserted, claim.supported, or claim.refuted.

Parameter	Type	Required	Description
`claimId`	string	Yes	Claim contribution `_id`
`action`	string	Yes	`upsert`, `support`, or `refute`
`text`	string	No	New claim text (for `upsert`)
`evidenceIds`	string[]	No	Evidence IDs (for `support`)
`refuterIds`	string[]	No	Refuting IDs (for `refute`)
`noveltyScore`	number	No	Updated novelty score

Status

`hubify_compute_status`

Current pod and compute status, cost totals, monthly budget, and budget headroom. No parameters.

`hubify_repo_status`

Lab GitHub repo wiring: URL, default branch, manifest schema version, auto-push state. No parameters.

Paper Pipeline

`hubify_get_paper_pipeline_status`

Return pipeline state for a paper: title, status, readiness %, target venue, last-updated.

Parameter	Type	Required	Description
`paperId`	string	Yes	Paper Convex `_id`

`hubify_update_paper_stage`

Move a paper to a new pipeline stage. Emits paper.stage.changed and updates the paper row.

Parameter	Type	Required	Description
`paperId`	string	Yes	Paper Convex `_id`
`from`	string	Yes	Current stage (e.g. `drafting`)
`to`	string	Yes	Target stage (e.g. `internal-review`)
`newStatus`	string	No	New top-level status if it should change: `draft`, `review`, `submitted`, `published`

`hubify_update_paper_section`

Record that a paper section was drafted, edited, or completed. Emits paper.section.drafted, paper.section.edited, or paper.section.completed. Note: does not write .tex content directly — the round-trip GitHub sync owns the file content.

Parameter	Type	Required	Description
`paperId`	string	Yes	Paper Convex `_id`
`section`	string	Yes	Section name (e.g. `introduction`, `methods`, `results`)
`action`	string	Yes	`draft`, `edit`, or `complete`
`sha`	string	No	Git commit sha if known
`summary`	string	No	Brief description of what changed

Adversarial Review

`hubify_start_adversarial_review`

Kick off a multi-model adversarial peer review for a paper. Emits adversarial_review.started and triggers parallel reviewer roles.

Parameter	Type	Required	Description
`paperId`	string	Yes	Paper Convex `_id`
`roles`	string[]	No	Reviewer roles (e.g. `["skeptic", "statistician"]`). Backend defaults apply if omitted.
`abstract`	string	No	Override the stored abstract for this review round

`hubify_log_model_review`

Record one model’s verdict in an in-flight adversarial review. Creates a paperReviews row visible in the reviews UI.

Parameter	Type	Required	Description
`paperId`	string	Yes	Paper Convex `_id`
`role`	string	Yes	Reviewer role (`skeptic`, `statistician`, `generalist`)
`model`	string	Yes	Model identifier (e.g. `claude-opus-4-8`)
`verdict`	string	Yes	`approve`, `concern`, or `reject`
`summary`	string	Yes	Review summary
`findings`	string[]	No	Specific findings
`correlationId`	string	No	Correlation ID from `hubify_start_adversarial_review`

`hubify_create_review_consensus`

Synthesize an adversarial review verdict. Emits consensus.created and triggers the synthesis action.

Parameter	Type	Required	Description
`paperId`	string	Yes	Paper Convex `_id`
`verdict`	string	Yes	`approve`, `minor`, `major`, or `reject`
`blockers`	string[]	No	Blocking issues that must be resolved
`correlationId`	string	No	Correlation ID from `hubify_start_adversarial_review`

MCP Resources

In addition to tools, the MCP server exposes read-only resources:

Resource URI	Contents
`lab://status`	Lab stats as JSON
`lab://experiments`	All experiments as JSON
`lab://papers`	All papers as JSON
`lab://agents`	All agents as JSON
`lab://tasks`	All tasks as JSON
`lab://memory`	Lab decisions + agent learnings as JSON

MCP Prompts

Four built-in prompt templates available in any MCP client:

Prompt	Description
`research-assistant`	Research assistant with full lab context injected
`paper-reviewer`	Rigorous peer reviewer for a specific paper
`experiment-designer`	Designs next experiments based on completed results
`houston-method`	Post-experiment 9-step Houston Method checklist
`standup-facilitator`	Facilitates agent standups with real lab data

​MCP Tools Reference

​Platform Tools

​Lab Navigation

​lab_create

​lab_status

​lab_list

​lab_switch

​Experiments

​experiment_list

​experiment_create

​experiment_run

​experiment_queue

​Papers

​paper_list

​paper_create

​paper_review

​Tasks

​task_list

​task_create

​task_update

​Memory & Knowledge

​memory_search

​memory_save

​knowledge_search

​knowledge_add

​agent_learning_save

​Contributions

​contribution_list

​contribution_create

​Agents & Communication

​agent_list

​chat

​Compute & Costs

​pod_list

​pod_create

​cost_summary

​Misc

​notification_list

​survey_list

​figure_list

​standup_trigger

​activity_list

​Phase E Ledger Tools (hubify_*)

​Context & Session

​hubify_get_context

​hubify_get_context_pack

​hubify_create_session_summary

​Event Logging

​hubify_log_event

​hubify_log_finding

​hubify_sync_events

​Run & Claims

​hubify_record_run

​hubify_create_claim

​hubify_update_claim

​Status

​hubify_compute_status

​hubify_repo_status

​Paper Pipeline

​hubify_get_paper_pipeline_status

​hubify_update_paper_stage

​hubify_update_paper_section

​Adversarial Review

​hubify_start_adversarial_review

​hubify_log_model_review

​hubify_create_review_consensus

​MCP Resources

​MCP Prompts