Configuring Memory

Fundamentally, memory is a way of adding persistent context to a chat buffer. CodeCompanion uses groups to create a collection of files that can be added to chats. Groups can also be linked to a parser which can offer post-processing customization such as parsing file paths and adding them as buffers or files to the chat buffer.

Enabling Memory

By default, memory is not enabled in the chat buffer. To enable it:

require("codecompanion").setup({
  memory = {
    opts = {
      chat = {
        enabled = true,
      },
    },
  },
})

Once enabled, the plugin will look to load a common, or default, set of files every time a chat buffer is created. Some of these files are:

• .cursorrules
• .goosehints
• .rules
• CLAUDE.md

You can also conditionally determine if memory should be added to a chat buffer:

require("codecompanion").setup({
  memory = {
    opts = {
      chat = {
        ---Function to determine if memory should be added to a chat buffer
        ---This requires `enabled` to be true
        ---@param chat CodeCompanion.Chat
        ---@return boolean
        condition = function(chat)
          return chat.adapter.type ~= "acp"
        end,
      }
    },
  },
})

The example above is taken from the config itself, as by default, the plugin will not add memory to ACP adapters.

Working with Groups

In the plugin, memory groups are a collection of files and/or directories. Below is an example of what a claude group might look like:

require("codecompanion").setup({
  memory = {
    claude = {
      description = "Memory files for Claude Code users",
      files = {
        "~/.claude/CLAUDE.md",
        "CLAUDE.md",
        "CLAUDE.local.md",
      },
    },
  },
})

You'll notice that the file paths can be local to the current working directory or point to an absolute location.

Conditionally Enabling Groups

You can also conditionally enable memory groups. For instance, the default CodeCompanion group has the following conditional:

require("codecompanion").setup({
  memory = {
    CodeCompanion = {
      description = "CodeCompanion plugin memory files",
      ---@return boolean
      enabled = function()
        -- Don't show this to users who aren't working on CodeCompanion itself
        return vim.fn.getcwd():find("codecompanion", 1, true) ~= nil
      end,
      files = {}, -- removed for brevity
    },
  },
})

Setting Parsers

You can also apply parsers at a group level, to ensure that all files in the group are parsed in the same way:

require("codecompanion").setup({
  memory = {
    claude = {
      description = "Memory files for Claude Code users",
      parser = "claude",
      files = {
        "CLAUDE.md",
        "CLAUDE.local.md",
        "~/.claude/CLAUDE.md",
      },
    },
  },
})

In the example above, every file within the group will be sent through the claude parser before being added to the chat buffer.

Alternatively, you can also apply parsers at a file level.

require("codecompanion").setup({
  memory = {
    default = {
      description = "Collection of common files for all projects",
      files = {
        { path = "CLAUDE.md", parser = "claude" },
        { path = "CLAUDE.local.md", parser = "claude" },
        { path = "~/.claude/CLAUDE.md", parser = "claude" },
      },
    },
  },
})

Or combine both approaches, whereby the parsers at a file level will take precedence.

Disabling Parsers

To disable a parser against a memory group, simply assign it a parser of none.

Nesting Groups

It's also possible to nest groups within a group. This can be a convenient way of applying the same conditional to multiple groups alongside keeping your config clean:

require("codecompanion").setup({
  memory = {
    CodeCompanion = {
      description = "CodeCompanion plugin memory files",
      parser = "claude",
      files = {
        ["acp"] = {
          description = "The ACP implementation",
          files = {
            ".codecompanion/acp/acp.md",
          },
        },
      },
    },
  },
})

In the example above, the main group is CodeCompanion and a sub-group, acp, sits within the files table. The claude parser sits across all of the groups.

When using the Action Palette or the slash command, the plugin will extract these nested groups and display them.

Changing Defaults

Groups

The plugin will look to load the default memory group by default. This can be changed by:

require("codecompanion").setup({
  memory = {
    opts = {
      chat = {
        default_memory = "my_new_group",
      },
    },
  },
})

Alternatively, you can select multiple groups:

require("codecompanion").setup({
  memory = {
    opts = {
      chat = {
        default_memory = { "default", "my_new_group" },
      },
    },
  },
})

Buffers and Files

If a parsed memory group contains links to files and they are Neovim buffers, you can set specific parameters, such as a pin or a watch:

require("codecompanion").setup({
  memory = {
    opts = {
      chat = {
        default_params = "watch", -- watch|pin - when adding a buffer to the chat
      },
    },
  },
})

NOTE

The claude parser has been specifically built to output linked files that take advantage of this.