CLI

aeqi is the command-line control surface for the Company OS.

Use it for three jobs:

• Hosted client - chat with agents and operate an existing Company runtime.
• MCP bridge - expose aeqi tools to Codex, Claude Code, and other AI clients.
• Private runtime - run a Company runtime with dashboard, API, scheduler, MCP socket, and agent execution in approved private deployments.

For hosted use, the CLI does not run the runtime on your laptop. It authenticates to the platform and talks to the managed Company runtime. For private deployments, aeqi start runs the approved runtime locally or on the customer's infrastructure.

You may still see trust in some secret keys, on-chain account fields, and protocol-level names. Treat it as the lower-level runtime/protocol selector behind the Company; the top-level command is aeqi company, not aeqi trust.

Surface Role

The CLI is the operator shell for the other surfaces:

The API is for software integration. MCP is for AI clients. The CLI is for operators and scripts.

Installation

The public path is hosted access. Accepted users receive CLI access through the hosted onboarding flow. Private deployments receive installation details during deployment scoping.

Mental model

The CLI has three jobs:

The most common hosted path is:

1. Create or join a Company in the dashboard.
2. Create keys in the dashboard.
3. Use aeqi chat when you want a terminal conversation with an existing agent.
4. Use aeqi mcp when you want Codex or Claude Code to work with Company memory, quests, agents, and code graph tools.

Hosted Company setup

Create two keys:

• Account key (ak_...) from Account -> API.
• Secret key (sk_...) from the Company API keys page.

For terminal chat:

export AEQI_API_KEY=ak_...
export AEQI_API_URL=https://app.aeqi.ai
aeqi chat

If your account has multiple Companies or roles, the CLI asks which one to use. You can also pin the context:

aeqi chat \
  --api-url https://app.aeqi.ai \
  --api-key ak_... \
  --entity <trust_runtime_id> \
  --agent "Executive Assistant"

aeqi chat lists your Companies, filters to your human roles when possible, lists active agents in the selected runtime, then opens a streaming session with the selected agent. Messages are sent to the hosted runtime over the platform API and WebSocket stream.

For MCP clients:

export AEQI_SECRET_KEY=sk_...
export AEQI_API_KEY=ak_...
export AEQI_PLATFORM_URL=https://app.aeqi.ai
aeqi mcp

Most users do not run aeqi mcp by hand. They configure Codex, Claude Code, or another MCP client to spawn it. See MCP.

Hosted user stories

Use the CLI like a terminal agent client

You already have a Company and an Executive Assistant agent. Run:

AEQI_API_KEY=ak_... AEQI_API_URL=https://app.aeqi.ai aeqi chat

Pick the Company, role, and agent. The CLI becomes a terminal chat surface for that runtime agent. The agent still runs inside aeqi, with its role, memory, tools, sessions, and event history. The CLI is only the transport.

Use Codex or Claude Code with the Company

Configure your MCP client to run aeqi mcp with AEQI_SECRET_KEY, AEQI_API_KEY, and AEQI_PLATFORM_URL.

The MCP client can then:

• call me to confirm which user and Company runtime it is operating as;
• search and store Ideas for durable memory;
• create, update, and close Quests;
• inspect and hire Agents;
• trigger or inspect Events;
• use the code graph before changing source.

This is the aeqi version of an AI coding workspace: the chat client supplies the interface, while aeqi supplies Company context, memory, work ledger, agents, and execution history.

Interact with existing agents

In a terminal, use aeqi chat --agent <name-or-id> to talk to one existing agent.

From MCP, use:

agents(action='list')
agents(action='get', agent='Executive Assistant')
quests(action='create', subject='Draft the launch checklist', agent='Executive Assistant')

Creating a Quest with an agent or agent_id delegates work to that runtime agent. Omitting the agent creates user-scoped work that is not automatically owned by one agent.

Create a new agent

From MCP, hire a runtime agent:

agents(action='hire', template='analyst')

Then delegate work:

quests(
  action='create',
  subject='Map the onboarding funnel',
  description='Review the current docs and produce a prioritized improvement plan.',
  agent='analyst'
)

The MCP client did not become the new agent. It asked the aeqi runtime to create one, then created work for that agent to execute.

Act as yourself, not as an agent

The normal hosted MCP connection acts as your account inside the selected Company runtime. AEQI_SECRET_KEY selects the runtime, and AEQI_API_KEY binds the request to your user account.

AEQI_AGENT is only a client hint for logs and context. It does not make the connection own Quests or impersonate a runtime agent. Pass agent or agent_id explicitly when you want to delegate to, inspect, or filter for an agent.

AEQI_AGENT_ID is reserved for cases where a client intentionally needs to bind to a specific runtime agent. Most human-operated CLI and MCP setups should leave it unset.

Private runtime commands

`aeqi init`

Initialize aeqi in the current directory. Lightweight — writes a stub config and leaves the rest to aeqi setup. Use this when you want to author a runtime config by hand instead of taking the wizard's choices.

aeqi init

`aeqi setup`

Bootstrap a local runtime. The wizard detects context:

• Inside a git repo: configures the current workspace.
• Outside a repo: writes to ~/.aeqi/.

aeqi setup
aeqi setup --runtime anthropic_agent
aeqi setup --service

Creates SQLite databases (aeqi.db, sessions.db), starter agents, and a default config. --service installs a per-user daemon service after bootstrap; --force overwrites existing starter files.

`aeqi start`

Run the local daemon, REST API, scheduler, dashboard, and runtime in one process. Default bind comes from config.

aeqi start
aeqi start --bind 127.0.0.1:8400

Hosted users usually do not run this. The hosted platform already runs the tenant runtime.

`aeqi daemon`

Manage a local daemon service.

aeqi daemon start
aeqi daemon status
aeqi daemon stop
aeqi daemon query status
aeqi daemon install --start
aeqi daemon uninstall --stop

`aeqi secrets`

Manage local provider keys and secrets.

aeqi secrets set OPENROUTER_API_KEY sk-or-...
aeqi secrets set ANTHROPIC_API_KEY sk-ant-...
aeqi secrets get OPENROUTER_API_KEY
aeqi secrets list
aeqi secrets delete OPENROUTER_API_KEY

`aeqi agent`

Manage local persistent agents.

aeqi agent list
aeqi agent spawn "Research Bot" --model anthropic/claude-sonnet-4.6
aeqi agent show "Research Bot"
aeqi agent registry
aeqi agent retire "Research Bot"
aeqi agent activate "Research Bot"

For hosted agent management from AI clients, prefer MCP agents(...).

`aeqi chat`

Open the interactive terminal chat. With AEQI_API_KEY set, it uses hosted account-key mode. Without it, it uses the local runtime config.

aeqi chat
aeqi chat --agent assistant
aeqi chat --entity <trust_runtime_id> --role <role_id> --agent <agent_id>

`aeqi mcp`

Run the MCP server over stdio. MCP clients spawn this process and call tools through it.

aeqi mcp

Hosted mode requires AEQI_SECRET_KEY, AEQI_API_KEY, and AEQI_PLATFORM_URL. Self-hosted mode uses the local runtime socket from the config.

`aeqi ideas`

Manage local Ideas.

aeqi ideas search "auth patterns"
aeqi ideas store "JWT refresh" "Use rotating refresh tokens with 7-day expiry"
aeqi ideas export --vault ~/notes/aeqi
aeqi ideas import --vault ~/notes/aeqi

`aeqi quests`

List and close local Quests. assign creates a Quest for a root agent.

aeqi assign "Write the onboarding checklist" --root assistant --priority high
aeqi ready
aeqi quests
aeqi quests --all
aeqi close <quest_id> --reason "shipped"

For hosted Quest operations from Codex or Claude Code, prefer MCP quests(...).

`aeqi graph`

Operate the local code intelligence graph.

aeqi graph index --root assistant
aeqi graph index --root assistant --full
aeqi graph stats --root assistant

`aeqi doctor`

Run diagnostics across the local install — config presence, database reachability, missing secrets, broken sockets.

aeqi doctor
aeqi doctor --fix       # auto-fix what can be auto-fixed
aeqi doctor --strict    # non-zero exit if any issues remain

`aeqi status`

One-shot snapshot of the local runtime state. Prints daemon health, active agents, recent quests, and config locations.

aeqi status

`aeqi monitor`

Consolidated operator monitor view. Either one-shot or streaming.

aeqi monitor
aeqi monitor --root assistant       # focus a single root agent
aeqi monitor --watch                # refresh continuously
aeqi monitor --watch --interval 5
aeqi monitor --json                 # emit as JSON for piping

`aeqi team`

Show system team and per-root teams. Useful for inspecting who works for whom inside a Company runtime.

aeqi team
aeqi team --root assistant

`aeqi company`

Company-kernel utilities for a local runtime. Two subcommands:

• aeqi company derive --entity-id <id> — derive the canonical Company identity for a Company/entity id (add --json for machine output).
• aeqi company root-only --runtime-dir <dir> --keep-company <id> — prune a hosted runtime so only one root Company/entity remains. Dry-runs by default; pass --apply (with --confirm <id>) to write.

aeqi company --help

`aeqi audit`

Query the decision audit trail. Filter by root agent, quest, or recency.

aeqi audit
aeqi audit --root assistant
aeqi audit --quest 67-123
aeqi audit --last 20

`aeqi config`

Reload the local configuration after editing aeqi.toml. Subcommands cover showing, validating, and refreshing the daemon's view of config.

aeqi config --help

`aeqi primer`

Emit the session-primer context the daemon would inject at session start. Used by Claude Code's session-start hook; not normally invoked by hand.

aeqi primer

Less-used commands

These exist but most operators won't reach for them:

• aeqi run — one-shot agent invocation with a prompt; useful for scripts.
• aeqi pipeline — explicit pipeline workflow operations.
• aeqi operation — track work across root agents.
• aeqi prompt — legacy prompt-template ops (the four primitives are agents/ideas/quests/events; prompt predates the renaming).
• aeqi hooks / aeqi hook — Claude Code hook helpers and worker pin.
• aeqi done — alternate quest-close path used by hook scripts.
• aeqi deps — suggest or auto-apply quest dependencies.
• aeqi web — start the web API server standalone (aeqi start is the usual entry point).
• aeqi seed — seed local databases for fresh installs.

Run aeqi <cmd> --help for the current flags.

Environment variables

Related

• MCP
• Authentication
• REST API
• IPC verbs
• Quickstart