io.github.contextium-io/mcp-server

Request access to a private library or workflow you cannot see, or — as the resource owner — list, approve, or deny pending requests. Use action="request" when a loaded workflow reports a locked library (see its requestAccess affordance). This works from the terminal — the user does not need to open the web app.

Add a comment to a file, or reply to an existing comment (pass parentCommentId). Authored as the connected user. Optionally anchor to a line with lineNumber. Replying to an external (guest) commenter also emails them.

Add a phase (work item) to a workflow's project plan. Phases are appended at the end of the existing list. If the workflow has no project plan yet, one is created automatically.

Add a task — autonomous by default. Just pass a title and the server will (a) auto-ROUTE the task to the workflow it best fits (matched from the task text against workflow names/descriptions), falling back to the workspace's "General" workflow when no workflow clearly relates, and (b) auto-ASSIGN it to the right team member(s) by role/default. Every task belongs to a workflow — there is no workspace-level task. The caller does NOT need to choose a workflow or assignee. Provide an explicit workflowId/workflowName and/or assigneeIds ONLY to override that decision (the user has already decided) — explicit values are used verbatim and skip auto-routing/auto-assignment for whatever was specified. Set autoRoute:false to skip AI routing and place the task directly in the "General" workflow. Tasks are stored as DB tasks under the chosen workflow by default; only legacy workflows that still have a file-backed project library store the task as a phase file tagged "task".

Autonomous one-shot project setup — creates a full Contextium workspace (library, agents, workflow, project-state.md) without asking any questions. Use when the user says "just set it up", "create a project for X", or is on mobile/voice.

Autonomous one-shot research workspace setup — creates a research library, agents, and workflow without asking any questions. Use when the user says "deep dive into X", "research X for me", or is on mobile/voice.

Use AI to break a broad task into 3-6 concrete, actionable subtasks. Claude analyses the task title and workflow context, then creates the subtasks automatically. Use when a task is too broad to complete in one step.

Capture a quick note or seed to remember later — saves into an existing project's seeds or a general Quick Notes library. Use ONLY for lightweight "remember this" notes: when the user says "remember this", "note that", "jot this down", or "add this to [project]". Do NOT use this for a brand-new business, product, app, SaaS, or venture idea ("I have an idea for…", "I want to start a business that…", "I want to build…") — those are not quick notes; route them to idea_kickoff, which runs a full business analysis. When unsure between capturing and kicking off a venture idea, prefer idea_kickoff.

Show all available Contextium capabilities and how to use them. Call this when the user asks: "what can Contextium do", "help with Contextium", "how do I use Contextium", "what Contextium tools are available", "show me what I can do with Contextium".

Show a Contextium workspace health dashboard — available workspaces, libraries, agents, skills, and workflows. Call this when the user asks: "check Contextium status", "what workspace am I in", "what resources do I have in Contextium", "show my Contextium setup", "what libraries do I have".

Create a new resource in a workspace. Set `type` to what you're creating: "file" (needs libraryId + title + complete content), "library" (a context library — needs name), or "agent" / "skill" / "workflow" (need name; a workflow can bundle projectIds, agentIds, skillIds, and tagIds). Always provide full, real content for files — never empty or placeholder. File linking: to link to another file in the SAME workspace, put //<fileId> inline in the content — use the target file's UUID (from search/list/get results), not its title. The link renders as the target's title and is indexed automatically on save; the target gains a backlink with no extra step. Add a link only where the text genuinely references an existing file; never invent //<id> for a file that does not exist (it becomes a dangling link). Use get({ type: "file_links", id }) to inspect a file's links and backlinks.

Create a new agent in the workspace. WHAT IS AN AGENT: An agent is a specialised AI assistant with a defined role, persona, and behavioural instructions. Agents are where you put "You are a..." system prompts, persona definitions, tone of voice rules, and task-specific instructions. Agents can have skills attached to them to give them access to reference knowledge. EXAMPLES of correct agent content: - "You are a senior customer support agent for Acme Corp. You handle billing and account queries..." - "You are a code reviewer specialising in Python. You check for security vulnerabilities..." DO NOT use create_agent to store reference material, policies, or documentation — use create_skill for that. IMPORTANT: You MUST always provide both a clear name AND a meaningful description — never leave description blank. The description should explain the agent's role, specialisation, and how it behaves. WORKFLOW ATTACHMENT: Always provide workflowId when creating an agent for a specific project or feature — this ensures the agent is immediately attached to the right workflow and won't be orphaned.

Creates a new identity in the workspace. Identities capture voice, tone, role, and audience context so AI produces work from the right perspective.

Create a phase plan for a workflow. A workflow can have at most one plan. Returns the new plan ID to use with add_phase. This is an alias of create_project.

Create a phase plan for a workflow (despite the name, this does NOT create a context library — it creates the project plan that holds phases). A workflow can have at most one plan. Returns the new plan ID to use with add_phase. Alias: create_plan.

Interactively create an agent, skill, or workflow in the current Contextium workspace. Call this when the user says things like: "create an agent", "make a new agent", "create a skill", "build a workflow", "I need a skill for X", "set up an agent that does X", "make an agent that can do X".

Delete a resource (moves to trash, restorable). type: file | skill | workflow

Permanently delete an agent from the workspace. This cannot be undone. System agents (harness agents) cannot be deleted.

Bulk delete all phases marked as done from a workflow's project plan. Use this to clean up completed work. Returns the number of phases deleted.

Permanently deletes an identity by ID. Use dryRun=true first to preview which workflows will have their identity_id set to null.

Delete a context library (moves to trash, recoverable within 30 days). Requires owner or admin role.

Delete a single phase from a project plan by its phase ID. Use this to remove a specific work item. To bulk-remove completed work, use delete_done_phases instead.

Dismiss a broadcast notification so it no longer appears in the inbox.

Dispatch an AI agent runner job with a natural-language intent. The runner autonomously selects the right agent, plans actions, and executes them. Returns a job ID you can poll with get_agent_job.

Update the title or description of an existing phase without changing its status or position. Use this to reword or clarify a phase.

Manage Figma connector integrations — list connected files or trigger a sync to pull design tokens, components, and page structure into your Contextium libraries.

Get full details of a resource by ID. Use workspace_activity for recent activity logs. Use activity_archive to read a specific month's archived logs by YYYY-MM (e.g. 2025-02). Use list_activity_archives to see which months have archives. Use workflow_activity for the activity digest since a workflow's last session. Use entity_activity for recent activity on a specific file or entity.

Get the status and result of a specific agent runner job. Poll this after dispatch_agent_job to check completion.

Gets the full profile of a single identity by ID, including voice descriptors, writing samples, and all contextual fields.

Get the notification inbox for the current user — both activity notifications and broadcasts.

Get the project plan attached to a workflow, including all its phases (open and completed). Returns phase titles, statuses, assigned agents, and sort order. Use this to understand what work is planned or in progress for a workflow.

Get the total number of unread notifications (activity + broadcasts) for the current user.

Autonomous business-idea kickoff. Call this FIRST and IMMEDIATELY whenever a user voices a brand-new business, product, app, SaaS, startup, or venture idea — e.g. "I have an idea for…", "I have a business idea", "I want to start a business that…", "I want to build a product that…", "what if there was an app that…", "here's a business idea", "idea: …". Returns instructions to run a FULL business analysis autonomously — SWOT, market size (live TAM/SAM/SOM), competitor analysis with funding/revenue/valuation, build plan, RICE scorecard, timeline, roles, two-tier startup financials, and a synthesised go/no-go verdict — written as files into a dedicated project library. Do NOT present a "what do you want to do with this idea" menu and do NOT run the step-by-step new_project wizard: the idea IS the go-ahead, so just do the research. Use new_project instead for a feature inside an EXISTING product; use capture_idea only for a lightweight "remember this" note.

List all resources of a given type in a workspace — the fast way to see what already exists before you act. Returns lightweight summaries (IDs, names, key fields), not full content: use `get` for one resource's full detail, or `search` to find resources by text or tag. Set `type` to choose what to list — workspaces, libraries (context libraries), agents, skills, files (pass libraryId), workflows, tags, file_tags (pass fileId), workspace_status, or members (with roles + activity). Uses your active workspace when workspaceId is omitted.

List recent agent runner jobs for the workspace — shows status (pending, running, done, failed), intent, and result summary.

Lists all agents (AI assistant configurations) in a workspace. Agents define assistant roles and capabilities.

List the comments on a file — open (unresolved) threads by default. Comments are feedback to act on: each shows the author (members and external share-link guests), the anchored line number if any, and the comment id needed by resolve_comment / add_comment replies.

Lists identities visible to the current user in a workspace. Identities capture voice, tone, role, and audience context so AI produces work from the right perspective.

Read lessons learned previously recorded in this workspace — including ones captured while working on OTHER projects. Call this before starting new work, when debugging something that smells familiar, or when the user asks what has been learned before. Optionally filter by project.

List all phases for a workflow's project plan, ordered by sort_order. Returns each phase with its ID, title, status (pending/in_progress/done/blocked), assigned agent name, and creation date. Use the phase IDs with update_phase_status to progress work.

List tasks. Omit workflowId to list ALL tasks across every workflow in the workspace. Provide workflowId to list a single workflow's tasks instead; these are DB tasks by default, or phase files tagged "task" for legacy file-backed workflows. Returns status, creator, priority, and due dates. Use sparingly — load_workflow already includes active workflow tasks.

List all context libraries currently in the workspace trash (deleted within the last 30 days, not yet permanently removed).

Load complete workflow context — agents, skills, files, and libraries. Use this to initialise a full workflow session.

Manage relationships: add/remove skills on agents (one at a time, or many in one call with apply_bulk/remove_bulk + skillIds), add/remove inline tags on files, and share/unshare a private workflow with an individual workspace member. For type=tag this EDITS THE FILE CONTENT — it writes (or deletes) a #<value> token in the file body, which is the source of truth for a file's tags; the tag is auto-created if new. (Equivalent to updating the file content yourself.) For type=workflow_member, add is an upsert (re-adding an existing member updates their role).

Guided, interactive project setup — walks the user through creating a Contextium workspace step by step, asking for confirmation at each stage. Use when the user is present and wants to work through the setup together.

Guided, interactive research workspace setup — walks the user through creating a research workspace step by step. Use when the user is present and wants to steer the research direction.

Mark one notification as read, or mark all notifications as read when no ID is provided.

Browse, inspect, and install shared skills, agents, and MCP servers from the Contextium Marketplace. Set `action`: "search" to browse listings (optionally filter by itemType skill/agent/mcp_server, category, or query); "get" for a listing's full details by slug; "install" to add a listing to a workspace by slug. Typically search first to find a slug, then get or install with it.

Start a guided new Contextium project setup. Call this tool FIRST whenever a user wants to set up a new project in Contextium, create libraries/agents/skills for a project, get started with Contextium, or organise their AI context. Returns complete step-by-step setup instructions that you must follow exactly.

Notify all eligible workspace team members about a resource you just created or updated — file, workflow, agent, or library. Use this when the user says things like "notify the team of this new file", "let everyone know about this workflow", or "notify @sarah about this". Respects each member's notification preferences and Do Not Disturb settings. Returns how many members were notified.