### Ahi Configuration File (ahi.yaml)

Source: https://github.com/upstash/ahi/blob/main/README.md

Example configuration for an Ahi project, defining tool paths, skill entrypoints, environment variables, setup commands, and agent definitions with schedules.

```yaml
tools: ./tools/
skills: ./skills/SKILL.md

env:
  BRAVE_API_KEY: # forwarded from local .env
  NODE_ENV: production # explicit value

setup:
  - npm install

agents:
  - name: my-agent
    harness: claude-code
    model: anthropic/claude-sonnet-4-6
    schedules:
      - cron: "0 9 * * *"
        prompt: "Do your daily task"
        timeout: 300000
        # optional; defaults to /workspace/home
        folder: /workspace/home
```

--------------------------------

### Install and Initialize Upstash AHI CLI

Source: https://github.com/upstash/ahi/blob/main/blog-agent-server.md

Installs the Upstash AHI CLI globally and initializes a new project.

```bash
npm install -g @upstash/ahi

ahi init
```

--------------------------------

### Ahi Installation and Usage

Source: https://github.com/upstash/ahi/blob/main/blog.md

Install the Ahi CLI globally and use commands to initialize a project, test agents locally, apply changes to the remote environment, and run agents remotely.

```bash
npm install -g @upstash/ahi
ahi init
ahi dev "do something useful"    # test locally
ahi apply                         # apply project state to the box
ahi run "do something useful"    # run remotely
```

--------------------------------

### Install Ahi CLI

Source: https://github.com/upstash/ahi/blob/main/README.md

Install the Ahi command-line interface globally using npm.

```bash
npm install -g @upstash/ahi
```

--------------------------------

### Ahi Project Structure Example

Source: https://github.com/upstash/ahi/blob/main/blog.md

Illustrates the conventional folder structure for an Ahi project, including tools, skills, and data directories.

```yaml
botstreet/
├── ahi.yaml
├── tools/
│   ├── trade.ts
│   ├── portfolio.ts
│   ├── prices.ts
│   ├── snapshot.ts
│   └── search.ts
├── skills/
│   └── SKILL.md
└── data/
    ├── portfolio.json
    └── memory.md
```

--------------------------------

### Ahi Quick Start Commands

Source: https://github.com/upstash/ahi/blob/main/README.md

Commands to initialize a new Ahi project, set up environment variables, run an agent locally, apply project state, and run an agent remotely.

```bash
ahi init                          # scaffold project structure
cp .env.example .env              # add your API keys
ahi dev "remember to buy milk"    # run agent locally
ahi apply                         # apply project state to the box
ahi run "list my notes"           # run remotely
```

--------------------------------

### Ahi Configuration Example

Source: https://github.com/upstash/ahi/blob/main/blog.md

Defines Ahi agents, specifying the paths for tools and skills, and mapping them to agents.

```yaml
tools: ./tools/
skills: ./skills/SKILL.md

agents:

```

--------------------------------

### Run an Agent with Upstash AHI CLI

Source: https://github.com/upstash/ahi/blob/main/blog-agent-server.md

Starts a development server for an agent with a specific task description.

```bash
ahi dev "Research the latest AI news and save today's summary"
```

--------------------------------

### Configure Multi-Model Agents

Source: https://context7.com/upstash/ahi/llms.txt

Set up multiple agents using different LLM models but sharing the same tools and skills. This configuration is for the Botstreet example.

```yaml
# ahi.yaml — three competing agents, same tools/skills, different models
tools: ./tools/
skills: ./skills/SKILL.md

agents:
  - name: botstreet-claude
    harness: claude-code
    model: anthropic/claude-opus-4-6
    schedules:
      - cron: "30 14 * * 1-5"
        prompt: "Run daily trading analysis, research market news, make trades, and save a snapshot"
        timeout: 600000

  - name: botstreet-gemini
    harness: claude-code
    model: openrouter/google/gemini-3.1-pro-preview
    schedules:
      - cron: "30 14 * * 1-5"
        prompt: "Run daily trading analysis, research market news, make trades, and save a snapshot"
        timeout: 600000

  - name: botstreet-openai
    harness: codex
    model: openai/gpt-5.4
    schedules:
      - cron: "30 14 * * 1-5"
        prompt: "Run daily trading analysis, research market news, make trades, and save a snapshot"
        timeout: 600000
```

--------------------------------

### Deploy Project Remotely (`ahi apply`)

Source: https://context7.com/upstash/ahi/llms.txt

Uploads project assets (tools, skills, env vars, setup commands, schedules) to remote agent boxes. Creates boxes if they don't exist. Local root files like `CLAUDE.md` and `AGENTS.md` are not uploaded.

```bash
# Deploy all agents defined in ahi.yaml
ahi apply

# Expected output:
# Applying local project to agent my-agent...
#   Using existing box box_abc123
# ✔ Uploaded 12 files
# ✔ Wrote 2 env var(s) to box
# ✔ Ran 1 setup command(s)
# ✔ Applied 1 schedule(s)
#   my-agent updated.
#
# Applied local project to all agents.

# Required environment variable:
export UPSTASH_BOX_API_KEY="your_key_here"
# Agent API keys are resolved by model prefix:
# anthropic/... → ANTHROPIC_API_KEY
# openai/...    → OPENAI_API_KEY
# openrouter/...→ OPENROUTER_API_KEY
# opencode/...  → OPENCODE_API_KEY
```

--------------------------------

### Run Daily News Researcher

Source: https://github.com/upstash/ahi/blob/main/examples/daily-news-researcher/README.md

Execute the daily news researcher project. This command starts the agent to research a topic and save a summary. Inspect the output file afterwards.

```bash
cd examples/daily-news-researcher
ahi dev "Research the latest AI news and save today's summary."
cat data/daily-summary.md
```

--------------------------------

### Botstreet Agent Configuration

Source: https://github.com/upstash/ahi/blob/main/README.md

Example YAML configuration for the Botstreet project, defining three agents with different models and schedules. Each agent is configured to run daily trading analysis.

```yaml
tools: ./tools/
skills: ./skills/SKILL.md

agents:
  - name: botstreet-claude
    harness: claude-code
    model: anthropic/claude-opus-4-6
    schedules:
      - cron: "30 14 * * 1-5"
        prompt: "Run daily trading analysis, research market news, make trades, and save a snapshot"
        timeout: 600000

  - name: botstreet-gemini
    harness: claude-code
    model: openrouter/google/gemini-3.1-pro-preview
    schedules:
      - cron: "30 14 * * 1-5"
        prompt: "Run daily trading analysis, research market news, make trades, and save a snapshot"
        timeout: 600000

  - name: botstreet-openai
    harness: codex
    model: openai/gpt-5.4
    schedules:
      - cron: "30 14 * * 1-5"
        prompt: "Run daily trading analysis, research market news, make trades, and save a snapshot"
        timeout: 600000
```

--------------------------------

### Run Ahi Dev Command

Source: https://github.com/upstash/ahi/blob/main/README.md

Executes an agent locally using the installed CLI. Use `--agent` to specify a particular agent from `ahi.yaml`.

```bash
ahi dev "save a note about the meeting"
```

```bash
ahi dev "list my notes" --agent my-agent
```

--------------------------------

### Run Agent Locally (`ahi dev`)

Source: https://context7.com/upstash/ahi/llms.txt

Runs an agent locally using the installed CLI. Automatically detects the harness based on available CLIs and root instruction files. It mirrors nested skill packages into the harness-native skill directory before launching.

```bash
# Basic local run — auto-detects harness from installed CLIs and root instruction files
ahi dev "remember to buy milk"

# Emulate a specific configured agent from ahi.yaml
ahi dev "list my notes" --agent my-agent

# Expected output:
# Running agent local locally
# Harness: claude-code | Model: local default
#
# <agent response streamed here>
```

--------------------------------

### Ahi Configuration Schema

Source: https://github.com/upstash/ahi/blob/main/plan.md

Defines the schema for the `ahi.yaml` configuration file, specifying tool paths, skill files, setup commands, and agent definitions.

```yaml
tools: ./tools/
skills: ./skills/SKILL.md
setup:
  - pnpm install --frozen-lockfile

agents:
  - name: agent-name
    model: claude-opus-4.6
```

--------------------------------

### Ahi Project Configuration (`ahi.yaml`)

Source: https://context7.com/upstash/ahi/llms.txt

The central configuration file for an Ahi project. Defines paths for tools and skills, environment variables, setup commands, and agent definitions including harness, model, and schedules.

```yaml
# ahi.yaml
tools: ./tools/
skills: ./skills/SKILL.md

env:
  BRAVE_API_KEY:          # forwarded from local .env (empty value = forward)
  NODE_ENV: production    # explicit value written directly to box

setup:
  - npm install           # runs on the box after each ahi apply

agents:
  - name: my-agent
    harness: claude-code                      # claude-code | codex | opencode
    model: anthropic/claude-sonnet-4-6        # anthropic/ | openai/ | openrouter/ | opencode/
    schedules:
      - cron: "0 9 * * *"
        prompt: "Do your daily task"
        timeout: 300000                       # milliseconds; optional
        folder: /workspace/home               # optional; defaults to /workspace/home

  - name: my-gpt-agent
    harness: codex
    model: openai/gpt-5.4

  - name: my-gemini-agent
    harness: claude-code
    model: openrouter/google/gemini-3.1-pro-preview
```

--------------------------------

### Ahi Config Fields

Source: https://github.com/upstash/ahi/blob/main/README.md

Key configuration fields for an Ahi project, specifying paths for tools, skills, environment variables, setup commands, and agent definitions.

```markdown
| Field    | Required | Description                                                                    |
| -------- | -------- | ------------------------------------------------------------------------------ |
| `tools`  | yes      | Path to tools directory                                                        |
| `skills` | yes      | Path to skill file                                                             |
| `env`    | no       | Env vars to write inside the box. Empty value = forward from local environment |
| `setup`  | no       | Commands to run on the box after file upload (e.g. `npm install`)              |
| `agents` | yes      | List of agent definitions                                                      |
```

--------------------------------

### Initialize New Ahi Project

Source: https://context7.com/upstash/ahi/llms.txt

Scaffolds a new Ahi project with a standard directory structure and sample files. Skips scaffolding if `ahi.yaml` already exists.

```bash
ahi init
# Output:
# Initialized Ahi project:
#   ahi.yaml
#   .env.example
#   CLAUDE.md
#   AGENTS.md
#   tools/note.ts
#   skills/SKILL.md
#   data/

# Resulting structure:
# my-project/
# ├── ahi.yaml          # agent and schedule definitions
# ├── CLAUDE.md         # local Claude Code development guidance
# ├── AGENTS.md         # local Codex/OpenCode development guidance
# ├── .env.example      # API key template
# ├── tools/note.ts     # sample tool script
# ├── skills/SKILL.md   # runtime skill (system prompt)
# └── data/             # durable agent state
```

--------------------------------

### Deploy Agent with Upstash Box SDK

Source: https://github.com/upstash/ahi/blob/main/blog-agent-server.md

Use the Box SDK to create an agent, upload files, and set up a schedule. This script also demonstrates how to run the agent immediately and stream its output.

```typescript
// deploy.ts
import { Box } from "@upstash/box";

const API_KEY = process.env.UPSTASH_BOX_API_KEY!;

async function deploy() {
  // Create a box with an agent
  const box = await Box.create({
    apiKey: API_KEY,
    name: "news-researcher",
    runtime: "node",
    agent: {
      provider: "claude",
      model: "claude-sonnet-4-6",
    },
  });

  // Upload the tool and skill
  await box.files.upload([
    { path: "tools/news-summary.ts", destination: "/workspace/home/tools/news-summary.ts" },
    { path: "skills/SKILL.md", destination: "/workspace/home/skills/SKILL.md" },
  ]);

  // Schedule a daily run at 9am UTC
  await box.schedule.agent({
    cron: "0 9 * * *",
    prompt: "Research the latest AI news and save today's one-sentence summary.",
    timeout: 300000,
  });

  console.log("Deployed. The agent will run every day at 9am UTC.");

  // Or run it right now
  const stream = await box.agent.stream({
    prompt: "Research the latest AI news and save today's one-sentence summary.",
  });

  for await (const chunk of stream) {
    if (chunk.type === "text-delta") {
      process.stdout.write(chunk.text);
    }
  }
}

deploy();

```

--------------------------------

### Recommended Ahi Skills Directory Structure

Source: https://github.com/upstash/ahi/blob/main/README.md

Illustrates a recommended structure for organizing multiple skill packages within the `skills/` directory. Nested skill packages are treated as reusable components.

```text
skills/
├── SKILL.md
├── notes/
│   ├── SKILL.md
│   └── templates/
└── research/
    ├── SKILL.md
    └── references/
```

--------------------------------

### Skill Package Structure

Source: https://context7.com/upstash/ahi/llms.txt

Illustrates a nested directory structure for organizing multiple agent skills, where each subdirectory contains a 'SKILL.md' file.

```text
skills/
├── SKILL.md           # primary router — reference nested packages
├── notes/
│   ├── SKILL.md       # notes skill package
│   └── templates/
└── research/
    ├── SKILL.md       # research skill package
    └── references/

```

--------------------------------

### Ahi Init Scaffold Structure

Source: https://github.com/upstash/ahi/blob/main/plan.md

Illustrates the default directory and file structure created by the `ahi init` command for a new Ahi project.

```markdown
my-project/
├── ahi.yaml          # single agent, placeholder model
├── .env.example      # example environment variables
├── tools/
│   └── note.ts       # simple durable note tool
├── skills/
│   └── SKILL.md      # starter note-keeper skill
└── data/             # empty, .gitkeep
```

--------------------------------

### Configure Environment Variables for Ahi

Source: https://context7.com/upstash/ahi/llms.txt

Set up necessary API keys for different LLM providers in a `.env` file. These variables are automatically loaded by Ahi commands.

```bash
# .env file — loaded automatically by all ahi commands
UPSTASH_BOX_API_KEY=upstash_box_xxx
ANTHROPIC_API_KEY=sk-ant-xxx
OPENAI_API_KEY=sk-xxx
OPENROUTER_API_KEY=sk-or-xxx
```

--------------------------------

### Ahi CLI Commands

Source: https://github.com/upstash/ahi/blob/main/blog.md

Scaffold project structure, run agents locally for fast iteration, execute agents remotely, apply project changes, and access the monitoring dashboard.

```bash
ahi init              # scaffold the folder structure
ahi dev "prompt"      # run an agent locally — fast iteration
ahi run "prompt"      # run an agent remotely on the box
ahi apply             # apply project files and schedules to the box
ahi console           # open the monitoring dashboard
```

--------------------------------

### Ahi CLI Commands Overview

Source: https://github.com/upstash/ahi/blob/main/plan.md

Lists the available commands for the Ahi CLI tool, including scaffolding, local development, remote execution, deployment, and console access.

```markdown
| Command | Description |
|---------|-------------|
| `ahi init` | Scaffold the folder structure (tools/, skills/, data/, ahi.yaml) |
| `ahi dev "prompt"` | Run an agent locally — fast iteration, no Box involved |
| `ahi run "prompt"` | Run an agent remotely on the Box, stream output back |
| `ahi apply` | Apply local project state to all agent boxes |
| `ahi console` | Open the monitoring dashboard |
```

--------------------------------

### Run Agent Prompt

Source: https://github.com/upstash/ahi/blob/main/README.md

Executes an agent with a given prompt on a remote box and streams the output. Optionally specify an agent name.

```bash
ahi run "analyze today's data"
ahi run "generate report" --agent my-agent
```

--------------------------------

### Box SDK Key Methods for Ahi

Source: https://github.com/upstash/ahi/blob/main/plan.md

Highlights essential methods from the `@upstash/box` SDK that are utilized by the Ahi framework for managing agents and their environments.

```markdown
- `Box.create(config)` — create a box with agent, runtime, env, skills, mcpServers
- `Box.get(boxId)` — reconnect to existing box
- `Box.list()` — list all boxes
- `box.agent.run({ prompt })` / `box.agent.stream({ prompt })` — run agent
- `box.fs.upload(files)` / `box.fs.list(path)` / `box.fs.read(path)` / `box.fs.download(files)
- `box.schedule.createPrompt(opts)` / `box.schedule.createExec(opts)` / `box.schedule.list()`
- `box.exec.command(cmd)` / `box.exec.stream(cmd)
- `box.pause()` / `box.resume()` / `box.delete()
- Auth: `UPSTASH_BOX_API_KEY` env var or `apiKey` param
```

--------------------------------

### Manage Multi-Model Agents with Ahi CLI

Source: https://context7.com/upstash/ahi/llms.txt

Commands for deploying multiple agents simultaneously and running prompts across them. Use `ahi apply` to deploy all agents.

```bash
# Deploy all three agents at once
ahi apply

# Run the same prompt across all agents in parallel
ahi run "summarize your portfolio performance"

# Pull data from a specific agent
ahi pull-data --agent botstreet-claude
```

--------------------------------

### Ahi CLI Commands

Source: https://github.com/upstash/ahi/blob/main/blog-agent-server.md

Manage agent deployments and runs using the Ahi CLI.

```bash
ahi apply   # Apply tools, skills, and schedules to the box
ahi run "Research the latest AI news"   # Run immediately
ahi console # Monitor runs, logs, files, schedules

```

--------------------------------

### Ahi CLI Configuration

Source: https://github.com/upstash/ahi/blob/main/blog-agent-server.md

Define agent projects using a folder structure and a YAML configuration file for tools, skills, and schedules.

```yaml
# ahi.yaml
tools: ./tools/
skills: ./skills/SKILL.md

agents:
  - name: news-researcher
    model: claude-sonnet-4-6
    schedules:
      - cron: "0 9 * * *"
        prompt: "Research the latest AI news and save today's one-sentence summary."
        timeout: 300000

```

--------------------------------

### Manage Scheduled Agents with Ahi CLI

Source: https://context7.com/upstash/ahi/llms.txt

Commands for local testing, deployment, and manual triggering of scheduled agents. Use `ahi dev` for local testing and `ahi apply` to deploy.

```bash
# Test locally before deploying
cd examples/daily-news-researcher
cp .env.example .env               # add ANTHROPIC_API_KEY
ahi dev "Research the latest AI news and save today's summary."
cat data/daily-summary.md

# Deploy to run on schedule
ahi apply

# Manually trigger a remote run
ahi run "Research the latest AI news and save today's summary."

# Pull saved summaries back locally
ahi pull-data
cat data/daily-summary.md
```

--------------------------------

### Configure a Daily News Research Agent

Source: https://context7.com/upstash/ahi/llms.txt

Define a scheduled agent that runs daily to research AI news and save a summary. Ensure ANTHROPIC_API_KEY is set in your environment.

```yaml
# examples/daily-news-researcher/ahi.yaml
tools: ./tools/
skills: ./skills/SKILL.md

agents:
  - name: daily-news-researcher
    harness: claude-code
    model: anthropic/claude-sonnet-4-6
    schedules:
      - cron: "0 14 * * *"           # runs every day at 14:00 UTC
        prompt: "Research the latest AI news and save today's one-sentence summary."
        timeout: 300000              # 5 minutes
```

--------------------------------

### Run Prompt on Remote Agent

Source: https://context7.com/upstash/ahi/llms.txt

Execute a prompt against a deployed agent on Upstash Box and stream output. Can target a specific agent or run on all agents in parallel.

```bash
ahi run "analyze today's data"
```

```bash
ahi run "generate report" --agent my-agent
```

```bash
ahi run "summarize your notes"
```

--------------------------------

### TypeScript Durable Notes Tool

Source: https://context7.com/upstash/ahi/llms.txt

A sample TypeScript tool for managing durable notes. It handles adding, listing, and clearing notes stored in a local file. Agents invoke this tool via shell commands.

```typescript
// tools/note.ts — sample durable notes tool
import { existsSync, mkdirSync, readFileSync, writeFileSync } from "fs";
import { resolve } from "path";

const command = process.argv[2];
const text = process.argv.slice(3).join(" ").trim();

const dataDir = resolve(process.cwd(), "data");
const notesPath = resolve(dataDir, "notes.md");

function ensureDataDir() {
  mkdirSync(dataDir, { recursive: true });
}

function readNotes(): string[] {
  if (!existsSync(notesPath)) return [];
  const content = readFileSync(notesPath, "utf8").trim();
  if (!content) return [];
  return content.split("\n").filter(Boolean).map((line) => line.replace(/^- /, ""));
}

function writeNotes(notes: string[]) {
  const content = notes.length > 0
    ? notes.map((note) => `- ${note}`).join("\n") + "\n"
    : "";
  writeFileSync(notesPath, content);
}

ensureDataDir();

switch (command) {
  case "add": {
    const notes = readNotes();
    notes.push(text);
    writeNotes(notes);
    console.log(`Saved note ${notes.length}: ${text}`);
    break;
  }
  case "list": {
    const notes = readNotes();
    if (notes.length === 0) { console.log("No notes saved."); break; }
    console.log(notes.map((note, i) => `${i + 1}. ${note}`).join("\n"));
    break;
  }
  case "clear":
    writeNotes([]);
    console.log("Cleared all notes.");
    break;
  default:
    console.log("Usage: npx tsx tools/note.ts <add|list|clear> [text]");
}

// Agent invokes tools via:
// npx tsx /workspace/home/tools/note.ts add "buy oat milk"
// npx tsx /workspace/home/tools/note.ts list
// npx tsx /workspace/home/tools/note.ts clear

```

--------------------------------

### List Saved Summaries with News Researcher Tool

Source: https://github.com/upstash/ahi/blob/main/examples/daily-news-researcher/skills/SKILL.md

Execute this command to retrieve a list of all previously saved news summaries.

```bash
npx tsx /workspace/home/tools/news-summary.ts list
```

--------------------------------

### News Summary Tool (TypeScript)

Source: https://github.com/upstash/ahi/blob/main/blog-agent-server.md

A TypeScript tool script for managing dated news summaries in a file. It uses Node.js built-ins and accepts commands like 'save', 'list', and 'latest' via process arguments.

```typescript
// tools/news-summary.ts
import { existsSync, mkdirSync, readFileSync, writeFileSync } from "fs";
import { resolve } from "path";

const command = process.argv[2];
const date = process.argv[3];
const summary = process.argv.slice(4).join(" ").trim();

const dataDir = resolve(process.cwd(), "data");
const summaryPath = resolve(dataDir, "daily-summary.md");

mkdirSync(dataDir, { recursive: true });

function readEntries(): string[] {
  if (!existsSync(summaryPath)) return [];
  const content = readFileSync(summaryPath, "utf8").trim();
  return content ? content.split("\n").filter(Boolean).map((l) => l.replace(/^- /, "")) : [];
}

function writeEntries(entries: string[]) {
  writeFileSync(summaryPath, entries.map((e) => `- ${e}`).join("\n") + "\n");
}

switch (command) {
  case "save": {
    const entries = readEntries();
    const entry = `${date}: ${summary}`;
    const existing = entries.findIndex((e) => e.startsWith(`${date}: `));
    if (existing >= 0) entries[existing] = entry;
    else entries.push(entry);
    writeEntries(entries);
    console.log(`Saved summary for ${date}.`);
    break;
  }
  case "list":
    console.log(readEntries().join("\n") || "No summaries saved.");
    break;
  case "latest": {
    const entries = readEntries();
    console.log(entries[entries.length - 1] || "No summaries saved.");
    break;
  }
  default:
    console.log("Usage: npx tsx tools/news-summary.ts <save|list|latest> [args]");
}

```

--------------------------------

### Ahi Skills Configuration

Source: https://github.com/upstash/ahi/blob/main/README.md

Defines the primary runtime skill file for an Ahi project. This path is the source of truth for deployed runtime behavior.

```yaml
skills: ./skills/SKILL.md
```

--------------------------------

### Push Agent Data

Source: https://github.com/upstash/ahi/blob/main/README.md

Uploads the local data directory to the agent's box. Supports specifying an agent or uploading all data.

```bash
ahi push-data
ahi push-data --agent my-agent
ahi push-data --all
```

--------------------------------

### Notes Skill Markdown

Source: https://context7.com/upstash/ahi/llms.txt

Defines a 'notes' skill for an agent, including its identity, available tools (add, list, clear), and processing logic. This Markdown file serves as the agent's system prompt.

```markdown
---
name: notes
description: Save, review, and clear simple notes for the user
---

# Notes Skill

## Identity

You are a simple note-keeping agent. Your job is to help the user save short
notes to durable storage, review what is already saved, and clear the list when asked.

## Tools

Save a note:
```
npx tsx /workspace/home/tools/note.ts add <text>
```

List saved notes:
```
npx tsx /workspace/home/tools/note.ts list
```

Clear all notes:
```
npx tsx /workspace/home/tools/note.ts clear
```

## Process

1. If the user wants to remember something, save it with the note tool.
2. If the user asks what is saved, list the notes first and then answer.
3. If the user asks to remove everything, use the clear command.
4. Keep responses short and explicit about what you saved or retrieved.

```

--------------------------------

### Daily News Researcher Skill (Markdown)

Source: https://github.com/upstash/ahi/blob/main/blog-agent-server.md

A Markdown file defining the 'daily-news-researcher' agent's identity, available tools, and operational process. It instructs the agent on how to research, summarize, and save news.

```markdown
---
name: daily-news-researcher
description: Research one news topic and save a one-sentence summary each day
---

# Daily News Researcher

## Identity

You are a daily news research agent.

Your job is to research the assigned topic, identify the most important
new development, and save a single concise summary sentence.

## Tools

Save a dated summary:
```
npx tsx /workspace/home/tools/news-summary.ts save <YYYY-MM-DD> <summary>
```

List saved summaries:
```
npx tsx /workspace/home/tools/news-summary.ts list
```

Show the latest saved summary:
```
npx tsx /workspace/home/tools/news-summary.ts latest
```

## Process

1. Research the requested topic using your available capabilities.
2. Focus on developments that are genuinely new for the current date.
3. Write exactly one sentence in plain language.
4. Save the sentence with today's date.
5. Don't duplicate a sentence if it's already saved for today.

```

--------------------------------

### Ahi Project Structure Convention

Source: https://github.com/upstash/ahi/blob/main/README.md

Standard directory layout for an Ahi project, including configuration, agent definitions, tools, skills, and data storage.

```plaintext
my-project/
├── ahi.yaml          # agent definitions
├── CLAUDE.md         # local Claude development guidance
├── AGENTS.md         # local Codex/OpenCode development guidance
├── .env              # API keys
├── tools/            # scripts agents execute
├── skills/           # reusable workflow instructions and native skill source
└── data/             # durable files managed by the agent
```

--------------------------------

### Show Latest Saved Summary with News Researcher Tool

Source: https://github.com/upstash/ahi/blob/main/examples/daily-news-researcher/skills/SKILL.md

Use this command to display the most recently saved news summary.

```bash
npx tsx /workspace/home/tools/news-summary.ts latest
```

--------------------------------

### Save Dated Summary with News Researcher Tool

Source: https://github.com/upstash/ahi/blob/main/examples/daily-news-researcher/skills/SKILL.md

Use this command to save a dated summary of a news topic. Ensure the date is in YYYY-MM-DD format and the summary is a single concise sentence.

```bash
npx tsx /workspace/home/tools/news-summary.ts save <YYYY-MM-DD> <summary>
```

--------------------------------

### Pull Agent Data

Source: https://github.com/upstash/ahi/blob/main/README.md

Downloads the agent's data directory from the box to the local project. Can be used with a specific agent name.

```bash
ahi pull-data
ahi pull-data --agent my-agent
```

--------------------------------

### Ahi Agent Fields

Source: https://github.com/upstash/ahi/blob/main/README.md

Defines the fields required for agent configurations within Ahi, including name, model, harness, and optional schedules.

```markdown
| Field       | Required | Description                                                              | 
| ----------- | -------- | ------------------------------------------------------------------------ |
| `name`      | yes      | Agent name (also the box name)                                           |
| `model`     | yes      | Model identifier (for example `anthropic/claude-sonnet-4-6`, `openai/gpt-5.4`, `openrouter/google/gemini-3.1-pro-preview`) |
| `harness`   | yes      | Runner to use for the agent: `claude-code`, `codex`, or `opencode`       |
| `schedules` | no       | Cron schedules with prompt, optional timeout, and optional folder override |
```

=== COMPLETE CONTENT === This response contains all available snippets from this library. No additional content exists. Do not make further requests.