MCP Server Documentation

ContextLayer MCP

Track, link, and recall context across your team — directly from Claude.

ContextLayer is a Model Context Protocol server that turns every conversation, task, and document into a typed knowledge graph the whole team can pull from. 62 tools across sessions, tasks, sprints, projects, teams, documents, and live activity — all backed by a unified intelligent context router.

Quickstart

ContextLayer is a remote MCP server running over Streamable HTTP. Connect from any MCP-compatible client (Claude Desktop, Claude.ai, Cursor, Cline, custom agents via the SDK).

Get a personal API token

Add the server to your MCP client

Endpoint: https://api.dotnova.io/mcp

Authentication: Authorization: Bearer <your_token>

claude_desktop_config.json

{
  "mcpServers": {
    "contextlayer": {
      "type": "http",
      "url": "https://api.dotnova.io/mcp",
      "headers": {
        "Authorization": "Bearer YOUR_TOKEN_HERE"
      }
    }
  }
}

Start using it

ContextLayer auto-starts a session on the first MCP call and silently tracks your work. Try: "What did our team do on the payments project this week?"

Authentication

All requests require a Bearer JWT in the Authorization header. Tokens are scoped to the user and the active organization, and inherit the user's role-based permissions (owner / admin / member).

Members see their own teams and projects.
Admins and owners see the whole organization.
Tokens can be revoked from the dashboard at any time.
The MCP endpoint also supports OAuth 2.0 discovery for clients that prefer it (returns WWW-Authenticate with the metadata URL).

Tools Reference

62 tools, grouped by domain. Every tool is fully self-describing via the MCP tools/list endpoint — your client will see the full input schema and description.

Discovery & Context (6)

Bootstrap a conversation with the user's profile, projects, and a unified context query that traverses the knowledge graph.

whoami	Profile, organization, accessible projects, recent sessions, top entities. Always call first.
get_context	Intelligent context retrieval — intent classification + graph traversal + semantic search returning critical/relevant/background tiers.
search_entities	Search people, companies, documents, concepts, and tools mentioned across the org.
get_my_stats	Personal activity statistics: sessions, events, active days, projects, recent activity.
get_org_stats	Org-wide stats: counts, engagement, task velocity (admin).
get_live_activity	What's happening right now — active sessions, events, traces, tasks in progress.

Sessions & Activities (7)

Sessions track work in real time and become typed knowledge blocks. Auto-started on the first MCP call; survives across multiple agents working on the same project.

start_session	Start (or reuse) a work session. Pass project_name to bind, or omit for a projectless session.
end_session	End a session. Triggers SessionBlock generation: activities are summarized into the knowledge graph.
log_activity	Log an event (code_change, decision, note, conversation, discovery). Auto-creates a session if none exists.
log_trace	Granular agent action (tool_call, file_read, file_write, code_search, api_call, thinking, error). High-frequency.
log_traces_batch	Log multiple traces in one call.
list_sessions	Recent work sessions, optionally filtered by project.
get_session_details	All events/activities in a session. Admins see any session in the org.

Tasks & Sprints (12)

Full task lifecycle, sprint planning, and dependency tracking. Tasks are linked to people, projects, and sessions in the graph.

create_task	Create a task with assignees, priority, tags, team, and due date.
list_tasks	Filter by project, status, sprint, or mine_only.
update_task	Move through workflow (todo → in_progress → done) or update fields.
delete_task	Remove a task permanently.
assign_task	Add a user to a task by name or email.
unassign_task	Remove a user from a task.
link_task_session	Connect a session to a task — feeds the knowledge graph.
add_task_dependency	Express that a task is blocked by another. Prevents cross-session contamination.
remove_task_dependency	Remove a dependency.
list_sprints	All sprints in a project: backlog, active, completed.
create_sprint	New sprint — auto-migrates non-completed tasks from the previous sprint.
move_task_to_sprint	Reassign a task to a different sprint.

Projects (9)

Project CRUD, membership, and team links.

list_projects	All projects accessible to the user.
create_project	New project — caller becomes owner.
update_project	Rename, change description, or update links.
delete_project	Permanently remove a project (admin).
get_project_status	Recent activity, members, and current state.
list_project_members	All people involved (via sessions and team links).
list_project_teams	Teams linked to a project.
link_team_project	Connect a team to a project.
unlink_team_project	Disconnect a team from a project.

Teams & People (8)

Team management, member roles, and people directory. Members see their own teams; admins see everyone.

list_teams	All teams in the organization.
get_team_details	Members, linked projects, and recent activity.
create_team	New team — caller becomes lead.
update_team	Rename or change description/objectives.
delete_team	Permanently remove a team (admin).
add_team_member	Add a user to a team by name or email.
remove_team_member	Remove a user from a team.
list_colleagues	People directory; filterable by team.

Knowledge Base & Documents (4)

Upload templates, policies, contracts, and references. Embeddings are generated automatically and surfaced as background context when relevant.

upload_document	Upload text or binary files (PDFs, images). Indexed and embedded automatically.
find_documents	Search by category, tags, project, or free text.
get_document	Read full content. Binary files come back base64-encoded.
delete_document	Permanently remove a document.

Organization & Users (8)

Organization setup, user management, and onboarding (admin/owner only).

setup_organization	Create teams, projects, and links in one step. Use during onboarding.
update_org	Org name, description, business model, objectives, custom AI context, links.
invite_user	Invite a new member with role, team, positions, and skills.
update_user	Change name, role, positions, or skills.
delete_user	Soft-delete a user from the organization.
remove_user_from_org	Remove a user without deleting the account.
list_pending_invites	Invites pending for the current user.
accept_org_invite	Accept a pending invite by org name.

Positions (4)

Job titles with member counts and permissions (admin only).

list_positions	All positions in the org.
create_position	Define a new job title.
update_position	Update title, description, or permissions.
delete_position	Remove a position permanently.

Context Routing (2)

Tune how the IntelligentRouter classifies queries.

get_context_routing	Current routing mode: keyword_llm, keyword_only, or llm_only.
set_context_routing	Change the routing mode (admin).

Use Cases

What a typical conversation with ContextLayer looks like:

Run sprint planning with full context

“Plan next sprint for the payments team — pull unfinished work, recent blockers, and team capacity, then draft the sprint goals.”

Generate status reports automatically

“Summarize what every team shipped this week and flag any project that's slipping against its deadline.”

Onboard new teammates

“Summarize everything the billing project has done in the last 60 days — decisions, key contributors, open tasks — for a new PM joining Monday.”

Coordinate across agents and tools

“Share context from my Claude Code session, my Cursor instance, and yesterday's research conversation so they all start with the same understanding of the auth refactor.”

Find internal experts and owners

“Who has the most experience with our Postgres migrations, and what have they recently shipped that I should review?”

Capture and link work automatically

“Log this planning conversation under the Q3 roadmap project, link it to the relevant epics, and create follow-up tasks for the action items we discussed.”

Data & Privacy

Data is processed and stored on servers in Germany (EU), hosted by Hetzner.
We collect only what's needed to operate the requested tools — no chat history, no Claude memory, no files from other contexts.
We do not sell data, share for behavioral advertising, or use your data to train AI models.
LGPD- and GDPR-compliant. Full policy: https://app.dotnova.io/privacy.
Data subject rights (access, correction, deletion, portability): email contato@dotnova.io — response within 15 days.

Support

Email: contato@dotnova.io
Dashboard: https://app.dotnova.io
Source: github.com/DotNova/context-layer
MCP registry: io.github.DotNova/context-layer