Files

Agency env file: one shared context file for all your agents

Every agency gets one shared "env" file (env.md) in the vault. Put any standing context in it, business facts, house rules, IDs, links, anything, and every agent, team, and the War Room can read it, and update it, over MCP. You can also read/write it from Claude Code.

The agency env file is a single free-form file, env.md: stored in your Vault, one per agency. Think of it as the shared "everything my agents should always know" note: your brand facts, key pipeline names, support email, house rules, IDs, links, whatever you want your AI to have on hand. It is scoped to your agency only; no other agency can see it.

What goes in it

Anything you like: it's plain text / markdown, not a rigid format. Common uses:

Brand facts (legal name, site, phone, support email).
Key GHL pipelines / stages your agents should target.
House rules ("never text before 9am", "always CC the office").
Reference IDs, account numbers, and links your agents need.

Note: it is not encrypted secret storage, treat it like a shared note your agents can read, not a password vault. For real API keys and secrets use Settings → Integrations.

Where to create / edit it

Go to Settings → Env File. It's a simple text box: type your content and click Save env file. That creates (or updates) env.md in your vault. That's the easiest way to set it up.

Other ways to read & update it

Floating assistant / War Room: just ask: "what's in my env file?" or "add X to my env file." The AI uses the env_read and env_write tools.
Any agent or team: they can call env_read to load your context and env_write to update it, both in chat/runs and over their published MCP endpoint.
Claude Code / claude.ai (via MCP): connect to an agent or team's MCP URL (from its Integrations tab). env_read and env_write show up as tools, so Claude Code can read your agency context and update it directly.
Files page: env.md is a normal vault file, so you can also open and edit it on the Files page.

The two tools

Tool	What it does
`env_read`	Returns the current env file content. If you haven't created one yet, it returns an empty result so the AI knows to offer to start one.
`env_write`	Creates or updates the env file. It replaces the whole file, so to add to it the AI reads it first, appends, then writes the whole thing back.

Isolation

The env file is strictly per-agency. Whichever surface reads or writes it, the agency is resolved on the server from your session (or the node's own agency for MCP), never from the AI, so your env file is only ever your agency's, and is never visible to another tenant.

## Agency env file: one shared context file for all your agents

Every agency gets one shared "env" file (env.md) in the vault. Put any standing context in it, business facts, house rules, IDs, links, anything, and every agent, team, and the War Room can read it, and update it, over MCP. You can also read/write it from Claude Code.

The **agency env file** is a single free-form file, `env.md`: stored in your Vault (/docs/file-library), one per agency. Think of it as the shared "everything my agents should always know" note: your brand facts, key pipeline names, support email, house rules, IDs, links, whatever you want your AI to have on hand. It is scoped to your agency only; no other agency can see it.

### What goes in it

Anything you like: it's plain text / markdown, not a rigid format. Common uses:

- Brand facts (legal name, site, phone, support email).

- Key GHL pipelines / stages your agents should target.

- House rules ("never text before 9am", "always CC the office").

- Reference IDs, account numbers, and links your agents need.

Note: it is not encrypted secret storage, treat it like a shared note your agents can read, not a password vault. For real API keys and secrets use Settings → Integrations (/docs/integrations).

### Where to create / edit it

Go to **Settings → Env File**. It's a simple text box: type your content and click Save env file. That creates (or updates) `env.md` in your vault. That's the easiest way to set it up.

### Other ways to read & update it

- Floating assistant / War Room: just ask: "what's in my env file?" or "add X to my env file." The AI uses the env_read and env_write tools.

- Any agent or team: they can call env_read to load your context and env_write to update it, both in chat/runs and over their published MCP endpoint.

- Claude Code / claude.ai (via MCP): connect to an agent or team's MCP URL (from its Integrations tab). env_read and env_write show up as tools, so Claude Code can read your agency context and update it directly.

- Files page: env.md is a normal vault file, so you can also open and edit it on the Files page.

### The two tools

| Tool | What it does |

| env_read | Returns the current env file content. If you haven't created one yet, it returns an empty result so the AI knows to offer to start one. |

| env_write | Creates or updates the env file. It replaces the whole file, so to add to it the AI reads it first, appends, then writes the whole thing back. |

### Isolation

The env file is strictly per-agency. Whichever surface reads or writes it, the agency is resolved on the server from your session (or the node's own agency for MCP), never from the AI, so your env file is only ever your agency's, and is never visible to another tenant.