Agent Insights

Agent Insights shows what your coding agents keep on disk: Claude Code’s permission rules, the memory and configuration files of your coding agents, and this project’s Claude Code conversations.

Open it from Sidebar › Agents › Agent Insights.

Header

The header shows the page title, a subtitle for the open view, and three tabs: Rules, Files and Conversations. The page opens on Rules.

The buttons on the right depend on the tab:

Tab	Buttons
Rules	Refresh reads the rule files again.
Files	A Read-only badge, Scan folder… and Add file….
Conversations	Refresh reads the conversation list again.

Rules

The Rules tab lists the Claude Code permission rules that apply in this project. They apply to every Claude Code session in the project, including all your Neova agents. Neova reads them from these files:

• ~/.claude/settings.json
• .claude/settings.json
• .claude/settings.local.json
• ~/.claude.json

The subtitle counts the rules and the files they come from.

Search and filters

Type in Search rules… to narrow the list. Matches are highlighted. Click the × to clear the search.

Below the search box, filter chips narrow the list further. Each chip shows how many rules it matches.

Filter	Shows
All rules	Every rule.
Added by “don’t ask again”	Rules in settings.local.json, the file Claude Code writes when you pick “don’t ask again”.
Denied	Deny rules.
Redundant	Rules another rule already covers, for example Bash(docker ps:*) under Bash(docker:*).
Global	Rules from files in your home folder.
Project	Rules from files in this project.

The table

Rules are grouped by category: Filesystem, Commands (Bash), Network (WebFetch), MCP servers & tools and Other tools. Each group header shows the number of rules and a short caption.

Column	Shows
Rule	The rule as written in the file. A covered by badge names the rule that already covers it.
Effect	Trusted, Denied, Ask or Allow.
Defined in	The file the rule comes from. An auto badge marks rules in settings.local.json.

Inside a group, restrictions come first. Allow rules for the same command, MCP server, domain or folder fold into one row with a chevron and a rule count. Click the row to show the rules inside it.

A long group shows the first rules and a Show N more link. Click Show less to fold it again. A search or the Redundant filter shows every rule.

How to remove a permission rule

1. Hover over the rule’s row.
2. Click the trash icon at the end of the row.

The rule is removed from its file and a message confirms it. Click Undo in that message to put the rule back. Claude Code picks up the change in running sessions and asks for the permission again the next time it needs it.

Hover over a folded row and click its trash icon to remove every rule in it.

When you remove a deny rule, Neova first asks Remove this deny rule?. Click Remove to go ahead or Keep it to cancel.

Some rows have a greyed-out trash icon. Hover over it to see why:

• Rules in ~/.claude.json are managed by Claude Code. Neova only reads that file.
• Some rows are settings, not rules. Turn them off in the file named in Defined in.

How to clean up redundant rules

1. Click the Redundant filter.
2. Check the list. Each rule shows which rule already covers it.
3. Click Remove all N on the right of the filter chips.

Files

The Files tab shows memory and configuration files of your coding agents, such as CLAUDE.md, AGENTS.md, settings.json or rules files. The files are read-only here. You choose which files appear.

On the left, files are grouped by coding agent: Claude Code, Cursor, Gemini, OpenCode, GitHub Copilot, Windsurf, Aider or Other. Under each agent, click Memory or Configuration to show those files.

The main area has two columns:

• Project: files inside this project.
• Global / External: files in your home folder or elsewhere.

Each file shows as a card with its label, its path, a scope badge (Project, Global or External), its format, and the file contents. Click the × on a card to remove it from this view. The file itself is not deleted.

How to add memory files

To find files automatically:

1. Click Scan folder… (or Scan a folder when the tab is empty).
2. Choose a folder, for example the project folder or your home folder.
3. In Scan results, tick the files you want.
4. Click Add N files.

To add one file by hand:

1. Click Add file… (or Add a file when the tab is empty).
2. Choose the file.
3. In Add memory or permissions file, set the fields below.
4. Click Add file.

Field	What it does
Label	The name on the card. Defaults to the file name.
Coding agent	The agent the file belongs to.
Type	Memory or Configuration.
Format	Markdown, JSON or Text.

Conversations

The Conversations tab shows this project’s Claude Code conversations. They are read-only.

The list on the left groups conversations under Today, Yesterday, This week and Earlier. Each row shows the title, when it was last updated and the git branch. A Live dot marks a conversation an agent terminal is in now.

Click a conversation to open it on the right. The header shows the title, when it started, the branch and the number of messages. Your messages appear on the right, Claude’s on the left. Tool calls show as short lines under Claude’s message. A conversation that is Live updates as the agent works. If you have scrolled up, a button shows how many new messages arrived. Click it to jump to the bottom.

How to search conversations

1. Type in Search conversations at the top of the list.
2. The list shows only conversations that contain the words, with a snippet of the match.
3. Open a conversation. A find bar appears in its header with the match count, for example 2 of 7.
4. Use the up and down arrows to step through the matches. Click the × to clear the search.

How to open an agent’s conversation from its terminal

1. In the agent’s terminal, click the ⋮ menu.
2. Click Open conversation.

Agent Insights opens on the Conversations tab with that conversation selected.

Related

• Sessions: the agent terminals with the Open conversation menu item.
• MCP Servers: the servers your agents can use in this project.