https://github.com/AIsa-team/cli

Primary Skill: Twitter Autopilot — demonstrated with Codex

The video presents AIsa as the transaction network for the agent economy: a single API key that gives an agent access to Skills, real-world APIs, and models when needed. The through-line is the before/after contrast: a brilliant but isolated agent that, with a single connection to AIsa, starts executing real tasks.

The centerpiece is a live demo of the Twitter Autopilot Skill running on Codex, followed by a walkthrough of the platform's three layers (APIs → Skills → Models) in that order.

Agreed framing: AIsa is NOT positioned only as a model provider. It is positioned as the resource and transaction layer that lets agents act in the real world. The model catalog is shown last, as a supporting layer.

2. Demo Setup (Codex + AIsa)

The demo runs entirely in Codex. The connection to AIsa uses the single-step flow recommended for custom agents, with no separate Twitter/X credentials to manage: OAuth is handled at the platform level by AIsa.

  1. Generate the API key in the AIsa console.

    export AISA_API_KEY="your-api-key-here"

  2. Connect Codex to AIsa by pasting the official configuration prompt:

    Read  and help me safely connect,
configure, and use AIsa's APIs, Skills, and LLMs in this agent environment.

  3. Install / invoke the Twitter Autopilot Skill, which runs on AISA_API_KEY.

Security note communicated on camera: Twitter/X credentials are never exposed in the prompt or inside the agent; the token lives at the platform level in AIsa. For new accounts, draft mode is used before publishing.

3. Video Structure (storyboard)

Sequence of blocks with what appears on screen and the idea conveyed in each.

Time On screen Message / idea conveyed
0:00–0:45 Cold open. Codex open; an agent that does not yet have access to live data. The pain: your agent is brilliant but isolated. What if a single key gave it APIs, Skills, and models?
0:45–2:30 AIsa website (home). Platform concept. What AIsa is: the transaction network for agents. One key = Skills + APIs + models. Not just a model provider.
2:30–7:30 LIVE demo in Codex with Twitter Autopilot. Three tasks (see section 4). The wow: the agent searches, monitors, and summarizes on X with no OAuth or developer accounts to set up.
7:30–9:30 Site walkthrough: /api → /skills → /models, in that order. The three layers: 1,000+ real APIs, the Skills library, and lastly the model catalog as a supporting layer.
9:30–11:00 Reveal of the "how": the Skill runs on AIsa with a single key. This Skill runs on AIsa. One platform, one key, every resource behind it.
11:00–12:00 /models section, brief mention of Chinese models. Pricing advantage: DeepSeek, Qwen, Kimi, GLM, MiniMax, ByteDance Seed at a fraction of the cost. Kept brief.
12:00–end Close + CTA. console.aisa.one on screen. Recap: from isolated to operational with one key. CTA: link in description + free credits.

4. Live Demo Detail (3 tasks)

The three tasks run in Codex with the Twitter Autopilot Skill. All are read/research or approved actions; nothing sensitive is published and no credentials are exposed. The prompts are shown on screen while the agent works. (On-screen prompts will be delivered in Spanish to match the video; English is shown here for the brand's review.)

Task 1 — Search and monitoring on X

  • On-screen prompt: "Search X for the latest on 'AI agents' and give me the most relevant tweets from the last 24 hours."
  • What it demonstrates: the agent accesses real-time X search through AIsa, with no developer account or Twitter API keys.
  • On screen: the agent calling the AIsa search endpoint and returning structured results.

Task 2 — Timeline summary / account monitoring

  • On-screen prompt: "Check the latest tweets from a reference account in the AI ecosystem and summarize what they're talking about and what changed."
  • What it demonstrates: competitor / key-account monitoring with automatic summarization, a use case the whole audience understands.
  • On screen: retrieval of the latest tweets + the agent-generated summary.

Task 3 — Reply draft / engagement (approval mode)

  • On-screen prompt: "Draft a reply to one of these tweets, in my style, and show it to me before publishing."
  • What it demonstrates: the publishing flow with OAuth handled by AIsa and a human-approval step (draft mode) before any write action.
  • On screen: the agent generating the draft and waiting for confirmation — emphasizing human control and safety.

Message to land (all 3 tasks): this Skill runs on AIsa. One platform and one key give the agent every resource behind the Skill, with no OAuth, token, or rate-limit juggling.

5. Must-Show Moments

Brand-requested elements that will explicitly appear in the video:

  • One key, every resource: the pain (before: developer accounts, OAuth, rate limits, scattered billing) and the solution (a single key) are conveyed.
  • Walkthrough of the three pages in order: [aisa.one/api]( → [aisa.one/skills]( → [aisa.one/models](
  • Recommended Skill demonstrated live: Twitter Autopilot, with OAuth handled by AIsa.
  • Chinese model advantage: brief mention of DeepSeek, Qwen, Kimi, GLM, MiniMax, and ByteDance Seed, focused on price. Comes after Skills and APIs.
  • Positioning: AIsa as the transaction network / resource layer for agents, not a simple model aggregator.

6. Close and Call to Action

  • Recap of the core message: from an isolated agent to an operational one with a single key.
  • CTA: link to [console.aisa.one]( in the description, mentioning free credits to get started.
  • Standard Fazt Tech channel close.

7. Approval

Please confirm: (a) the flow of the three demo tasks, (b) the order of the APIs → Skills → Models layers, (c) the positioning framing, and (d) any adjustments to messaging or elements to highlight.