# AI Agent Task Execution with Tool Orchestration

> AI Agent Task Execution with Tool Orchestration is a paid API for AI agents from agentsvc.io, paid per call via x402, $0.5/call, status unknown (last checked 2026-10-16).

Executes a multi-step research or lookup task using a Claude Haiku agent that autonomously plans and calls up to 38 specialized tools in parallel, returning a Markdown answer with sources.

## Facts

- Endpoint: POST https://agentsvc.io/api/v1/proxy/agent-run?utm_source=zero.xyz
- Price: $0.5/call
- Payment: x402
- Status: unknown
- Last checked: 2026-10-16
- Activations on Zero: 0
- Tags: x402
- Canonical page: https://www.zero.xyz/c/ai-agent-task-execution-with-tool-orchestration-76daa8e4
- Structured record (JSON): https://api.zero.xyz/v1/capabilities/cap_XHHBtksPwhrlasu3XeI6T

Status and success rate cover calls made through Zero and Zero's own probes. Third-party monitors may report differently.

## How to call it through Zero

Zero handles the 402 payment challenge and records the run. With the Zero CLI installed (`npm i -g @zeroxyz/cli`):

```sh
zero fetch --capability ai-agent-task-execution-with-tool-orchestration-76daa8e4 -d '<json body>'
```

Example prompt: Research Aave vs Morpho for me — get their current TVL, top yield opportunities, and any recent news — and give me a Markdown summary with sources. Use up to 10 steps.

## When to prefer this

Choose this endpoint when you need to answer a multi-faceted question that requires combining data from many sources (web, news, DeFi, crypto, DNS, compliance, etc.) in a single call, and you don't want to wire up individual tool calls yourself. It's ideal for open-ended research tasks, due diligence, pre-trade checks, or domain audits where the planning and tool selection should be automatic. Prefer a more targeted single-tool endpoint if your task is narrow and latency-sensitive, since this endpoint takes up to 90 seconds.

## Known failure modes

- Task too vague or ambiguous — agent may return a low-quality or off-topic answer
- max_steps too low to complete a complex task — answer may be incomplete
- Tool failures mid-run — individual tool calls may fail (ok: false) but agent continues with partial data
- Rate limits or payment failure — returns 402 if USDC credit is insufficient
- Timeout — tasks exceeding ~90 seconds may be cut short
- Hallucination risk — answer is AI-generated from tool outputs; sources should be verified

## How this service works

Hand off a whole research or lookup task in one call: a Claude Haiku 5.5 agent plans the steps, calls up to 38 agentsvc.io tools in parallel (web and news search, page reading, papers, GitHub and npm, company registers and sanctions, crypto prices, token checks, DeFi TVL and yields, wallets, FX and macro data, weather, DNS, WHOIS, SSL, CVEs) and returns a Markdown answer with sources and the list of tool calls it made. Up to 12 steps and about 90 seconds.

## Output

A JSON object containing a Markdown-formatted answer with inline source citations, a list of tool calls made (each with tool name, input, latency, and success status), model usage stats (tokens in/out, seconds elapsed, tool call count), and the model identifier used. Includes a caveat note that the answer is AI-written from tool data.

## Request schema (JSON Schema)

```json
{
 "type": "object",
 "properties": {
  "task": {
   "type": "string",
   "description": "What to find out or do, in plain words, max 4000 characters"
  },
  "max_steps": {
   "type": "integer",
   "default": 8,
   "maximum": 12,
   "minimum": 1,
   "description": "Upper limit of model rounds"
  }
 }
}
```

## Response schema (JSON Schema)

```json
{
 "type": "json",
 "example": {
  "data": {
   "note": "AI-written from tool data; check the sources before acting on it.",
   "model": "claude-haiku-5-5",
   "steps": [
    {
     "ms": 460,
     "ok": true,
     "tool": "defi-tvl",
     "input": {
      "mode": "protocol",
      "protocol": "aave"
     }
    },
    {
     "ms": 617,
     "ok": true,
     "tool": "defi-tvl",
     "input": {
      "mode": "protocol",
      "protocol": "morpho"
     }
    }
   ],
   "usage": {
    "seconds": 3.7,
    "tool_calls": 2,
    "model_input_tokens": 28176,
    "model_output_tokens": 397
   },
   "tools_used": [
    "defi-tvl"
   ],
   "answer_markdown": "Aave's TVL is about $16.8B on Aave V3 (the largest version, with $13.9B on Ethereum), plus about $0.65B on Aave V4 and $0.26B on Aave Horizon RWA, for roughly $17.7B in total across Aave's lending deployments. Morpho's TVL is about $11.0B on Morpho Blue (Ethereum $5.0B, Base $4.4B) plus about $7.4M on Morpho Midnight, as of DefiLlama's 2026-10-15 10:32 UTC snapshot.\n\nSources:\n- DefiLlama via defi-tvl tool, protocol mode: aave\n- DefiLlama via defi-tvl tool, protocol mode: morpho"
  },
  "success": true
 }
}
```

## More

- Live health (JSON, refreshed every minute): https://www.zero.xyz/c/ai-agent-task-execution-with-tool-orchestration-76daa8e4/health.json
- [Zero catalog index](https://www.zero.xyz/llms.txt)
- [Other services from agentsvc.io](https://www.zero.xyz/host/agentsvc.io/llms.txt)