Claude Code Integration

Connect Claude Code to your Cognova workspace for persistent memory, task management, agent queries, and knowledge access. Available as a plugin (with automated hooks) or a cross-agent skill.

Plan Requirements

All tools are available on every plan, including Free. Agent invocation (ask_agent) consumes credits.

Install

Plugin (recommended)

The Cognova plugin includes automated hooks that make your workspace proactive:

Session start — silently recalls memories related to your current project
Task creation — checks for duplicates before creating

/plugin marketplace add cognova-dev/cognova-skill
/plugin install cognova@cognova-dev

Then run /cognova to configure your API key.

Cross-Agent Skill (skills.sh)

Works in Claude Code, Cursor, Codex CLI, Gemini CLI, GitHub Copilot, and 40+ other agents. Installs the skill only (no hooks).

npx skills add cognova-dev/cognova-skill

Manual MCP

If you just want the tools without the skill or hooks:

claude mcp add cognova https://cognova.dev/mcp \
  --transport http \
  --header "Authorization: Bearer cg_YOUR_KEY"

What You Get

Once connected, Claude Code can use cognova_* tools directly:

Tasks & Projects

cognova_list_tasks          — List tasks (filter by project, status)
cognova_create_task         — Create a task
cognova_complete_task       — Mark a task as done
cognova_list_projects       — List all projects
cognova_create_project      — Create a project

AI Agents

cognova_list_agents         — List workspace agents
cognova_ask_agent           — Ask an agent a question (uses credits)

Memory & Knowledge

cognova_recall              — Search stored memories
cognova_remember            — Store a memory (decisions, patterns, facts)
cognova_list_knowledge      — List knowledge files
cognova_get_knowledge       — Read a knowledge file
cognova_create_knowledge    — Create a new knowledge file
cognova_update_knowledge    — Update a knowledge file's content

Secrets & Billing

cognova_list_secrets        — List secret key names
cognova_get_secret          — Get decrypted secret value
cognova_set_secret          — Store an encrypted secret
cognova_get_balance         — Check credit balance and plan tier

Documentation

cognova_search_docs         — Search platform docs

Plugin vs Skill vs MCP

Feature	Plugin	Skill	MCP Only
Auto-recall memories on session start	Yes	No	No
Auto-store decisions/findings (via skill instructions)	Yes	Yes	No
Duplicate task prevention	Yes	No	No
Tool discovery	Automatic	Automatic	Automatic
Cross-agent support	Claude Code only	40+ agents	Any MCP client
Setup	2 commands	1 command	1 command

The plugin provides the richest experience for Claude Code users. The skill works everywhere. MCP is the lightest option.

CLAUDE.md Snippet

To reinforce Cognova usage in any project, add this to your CLAUDE.md:

## Cognova Integration

This project uses Cognova for task management, knowledge, and agent queries.
MCP server configured — use cognova_* tools for workspace interaction.
Proactively store design decisions and findings as memories with cognova_remember.
Search memories with cognova_recall before asking about project context.

Source

The skill and plugin are open source: github.com/cognova-dev/cognova-skill