Algobot Cli

Algobot CLI

Algolia Agent Studio is Algolia's platform for building AI agents — RAG systems, conversational experiences, genAI content generation — with Algolia search and recommendations as the retrieval backbone. algobot (npm install -g algobot-ai) is its CLI for creating, testing, and deploying agents.

When to Use Algobot vs Other Tools

| Need | Use | |------|-----| | Build/manage AI agents on Algolia (RAG, chatbot, genAI UI) | algobot-cli (this skill) | | Algolia search index ops (records, settings, synonyms) | algolia-cli | | Search queries, analytics, recommendations | algolia-mcp |

Setup

npm install -g algobot-ai
algobot init                    # Interactive wizard (creates first agent + profile)

Or add a profile manually (CI-safe, non-interactive):

algobot profiles add --name prod --env prod

Non-Interactive Mode (Critical for Agents)

The TUI won't render in non-TTY environments (CI, scripts, agent subprocesses). Use these instead:

algobot ask "What is your return policy?"                    # One-shot
algobot interactive --text "hello ||| list my orders ||| /context"  # Multi-turn
algobot agents list --jq '.[] | .name'                       # JSON filtering

--jq is built-in — no need to install jq separately.

Core Commands

Agent Management

algobot agents list
algobot agents get 
algobot agents create --name "Support Bot" --model gpt-4o
algobot agents update  --name "New Name"
algobot agents publish 
algobot agents unpublish 
algobot agents delete 
algobot agents copy  --from-env dev --to-env prod

Chatting with an Agent

algobot ask "Find wireless headphones under $100"
algobot --profile staging ask "hello"        # Target specific environment
algobot --verbose ask "debug this"           # Show full HTTP traces

Profile / Environment Management

algobot profiles list
algobot profiles add --name dev --env dev
algobot profiles setdefault prod
algobot --profile prod agents list
algobot --env dev agents list

Config-as-Code Workflow

Version-control agent definitions with mustache templates — ideal for repeatable deployments across events, teams, or environments.

# 1. Scaffold from an existing agent
algobot agents scaffold            # → agent-config.json + PROMPT.md

# 2. Preview (API-level kill switch — mutations blocked by construction)
algobot --dry-run agents create --config agent-config.json --var event="Spring 2026"

# 3. Deploy
algobot agents create --config agent-config.json --var event_name="Spring 2026"

# 4. Update + publish in one step
algobot agents update  --config agent-config.json --var event_name="Summer 2026" --publish

{{key}} in JSON fields: JSON-safe escaping. In .md instructions: raw substitution.

Agent Studio Capabilities (via agent config)

Beyond basic chat, Agent Studio agents support:

• Tools: Algolia Search, Algolia Browse, Algolia Recommend, client-side tools, and MCP tools (connect CRMs, inventory systems, external APIs alongside Algolia). Manage with algobot tools list/add/remove
• Memory: Semantic (facts/preferences) and episodic (past interactions) memory across sessions, using algolia_memorize, algolia_ponder, and algolia_memory_search tools. Configure retrieval mode (preload vs preflight) in agent config.
• Conversation storage: Persistent history with configurable retention — see algobot conversations for export/delete
• Experimental: Citation markers [1][2] on responses, date injection, response caching — enable in agent config

Use algobot agents scaffold to inspect/edit these settings, algobot --dry-run to preview before applying.

Live Development

algobot agents watch patch.json             # Auto-apply patches on file change

Global Flags

| Flag | Effect | |------|--------| | --env dev\|staging\|prod\|local | Target environment | | --profile | Use named profile | | --dry-run | Preview without mutating (API-enforced) | | --verbose | Full HTTP logs | | --jq '' | Filter JSON output | | --confirm | Skip exec tool confirmations |

Gotchas

• TUI requires TTY. algobot with no args launches the TUI — hangs in scripts. Always use ask or --text in non-interactive contexts.
• Exit codes are 0/1 only in v2.0. Can't distinguish "not found" from "auth error" — parse stderr if needed.
• --output json missing on most commands in v2.0. Use --jq or JSON-structured stdout.
• algobot init is interactive. Don't use in CI — use profiles add with flags instead.
• Auth stored in ~/.algobot-cookie (AES-256-GCM). Inspect with algobot auth show.
• --config auto-discovers agent-config.json in cwd. Explicit: --config path/to/config.json.
• algobot = dev/deploy tool; REST API = production invocation. Use algobot to build and publish agents; call the Agent Studio completions API directly from your app. Don't guess the endpoint URL — run algobot agents get to retrieve it, or check the Agent Studio dashboard.
• Completions URL has no /agent-studio prefix. Use agent-studio..algolia.com/1/agents//completions?stream=false&compatibilityMode=ai-sdk-4. compatibilityMode is required (omitting → 422). Response: { id, role, content } with content as a flat string.
• agents create --provider does not link the provider. Agent stays draft, publish 409s. After create, run algobot agents patch --json patch.json with {"providerId": ""}, then publish.

Reference Docs

• [Command Reference](references/commands.md) — Full flags for every command
• [Config-as-Code Guide](references/config-as-code.md) — Templates, variables, multi-env patterns

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: algolia
• Source: algolia/skills
• License: MIT

Install and usage instructions live in the source repository linked above.