Web Search Plus for Hermes
Give Hermes multi-provider web search and clean extraction backed by real source links.
- Source repo
- robbyczgw-cla/hermes-web-search-plus
- Stars
- ★ 445
- Last updated
- 10d ago
- License
- MIT
- Primary language
- Python
- FA score
- 78/100 · Good
At a glance
- How it runs
- Works with
- Platform-specific
- Cost
- Free, no paid service needed
- Setup effort
- Medium · a few setup steps
- You'll need
- Typical use
- A Hermes researcher following recent events who needs every search result tied to an inspectable source page.
- Not a fit if
- Teams that do not use Hermes Agent
- Users who require completely offline research
- Teams requiring every provider to participate in automatic routing
- Website
- websearchplus.xyz ↗ set on GitHub
- Maintenance
- Commits in the last 30 dayslast commit 2026-11-10latest release v5.1.9 (2026-11-10)
- Source review
- 78/100 · Good
What does this agent do, and when should you use it?
Web Search Plus is a web-search and page-extraction plugin for the Hermes Agent plugin API, with v4.3.3 identified as the current release. It registers two tools: web_search_plus discovers results with original links, while web_extract_plus cleans content from URLs the agent already has. Its engine can route across multiple providers, fall back after failures or empty results, and use options such as SearXNG, Keenable, or a separately installed DonSeTch process. The core plugin requires Python 3.10+, installs through Hermes, and stores provider credentials in the active Hermes environment file. DonSeTch runs outside the package as a stdio MCP process, while Jev, Octen, and TinyFish are separately enabled integrations with documented boundaries.
A web_search_plus call accepts a query plus optional provider, freshness, locale, domain, cache, and research-mode controls. It calls configured search services and returns source records containing titles, snippets, and original URLs; automatic routing can try another eligible provider when one fails or returns nothing. The routing system can collect provider samples in provider_stats.json, and its cache exposes no_cache and cache_ttl. Given URLs, web_extract_plus validates them, invokes an extraction provider, and produces cleaned page text; the HTTP path enforces response-size limits and same-origin redirects. The optional DonSeTch adapter launches a separately installed stdio MCP executable selected through DONSETCH_BIN for explicit search or Markdown extraction. Operational interfaces include the setup.py configuration and status commands, a search CLI for smoke tests, and an optional wsp backend that redirects Hermes' native web_search and web_extract calls through the same engine.
- A Hermes researcher following recent events who needs every search result tied to an inspectable source page.
- An analyst with an existing URL list who wants cleaned page bodies that an agent can read more easily.
- A team exposed to search-API outages that wants controlled routing and fallback across several configured providers.
- An organization operating SearXNG that wants to combine it with keyless Keenable and reduce reliance on commercial APIs.
- A Hermes user seeking local source search or Markdown extraction who is willing to install and explicitly select DonSeTch.
- A plugin operator diagnosing retrieval behavior through status checks, cache controls, and quality reports.
Key setup facts
Pulled from the install and usage notes below. The project README has the current values.
- Install / run
npm install -g [email protected]
- Environment variables
MONID_API_KEYTINYFISH_API_KEY- Config files
~/.hermes/plugins/web-search-plus/setup.pyconfig.json~/.hermes/plugins/web-search-plus- Runtime
- Python 3.10+
How do you install or deploy this agent?
Hermes Agent and Python 3.10+ are required. Install and enable the plugin first:
hermes plugins install robbyczgw-cla/hermes-web-search-plus --enableInspect readiness and run the starter configuration:
python3 ~/.hermes/plugins/web-search-plus/setup.py status
python3 ~/.hermes/plugins/web-search-plus/setup.py setup --preset starterConfigure at least one search-capable key or local endpoint; extraction likewise requires an extraction-capable key or endpoint. The helper writes keys to the active Hermes environment file rather than config.json. Restart the Hermes CLI, or issue /reset in the session; Gateway users should run /restart followed by /reset.
For a setup without commercial API keys, use the self-hosted preset with your own SearXNG instance. It also enables Keenable's public extraction tier:
python3 ~/.hermes/plugins/web-search-plus/setup.py setup --preset self-hosted
python3 ~/.hermes/plugins/web-search-plus/setup.py statusDonSeTch is not bundled. Install the tested version separately and set DONSETCH_BIN to its absolute executable path if that adapter is needed:
npm install -g [email protected]How do you use this agent?
After reloading Hermes, call the two registered tools directly. Search while retaining source links:
web_search_plus(query="Hermes Agent latest release")
web_search_plus(query="compare recent open-source agent frameworks", mode="research", quality_report=True)Extract clean text from URLs already collected:
web_extract_plus(urls=["https://example.com"])
web_extract_plus(urls=["https://docs.example.com"], provider="linkup", render_js=False)A shell smoke test is also documented:
cd ~/.hermes/plugins/web-search-plus
python3 search.py --query "Hermes Agent latest release" --provider auto --quality-reportOcten, TinyFish, and DonSeTch stay outside automatic routing and fallback by default, so select them explicitly. Octen requires MONID_API_KEY:
web_search_plus(query="recent vector database research", provider="octen", freshness="month")TinyFish requires the user's own TINYFISH_API_KEY:
web_search_plus(query="recent agent framework releases", provider="tinyfish", freshness="week")What are this agent's strengths and limitations?
- Unifies 15 search providers and nine extraction providers behind two Hermes tools, while requiring only one suitable provider to activate each tool.
- Returns original source URLs instead of hiding the web behind a generated answer, which suits verification-heavy research.
- Combines automatic routing, failure fallback, cache controls, quality reports, and provider statistics to reduce single-service dead ends.
- Offers SearXNG, Keenable, and separate DonSeTch paths that can reduce dependence on paid search APIs.
- Hardens extraction with strict URL validation, same-origin redirects, response-size limits, and an untrusted-web-data notice.
- The core delivery target is Hermes Agent; the source documents no native ChatGPT, Codex, Claude, OpenAI API, or Claude API integration.
- Useful operation still depends on external search or extraction endpoints; the self-hosted profile needs SearXNG, and Keenable's public tier is rate-limited with no SLA.
- Provider privacy terms vary; TinyFish's standard terms permit Customer Data to be used for model training, requiring care with sensitive queries.
- DonSeTch, Jev, and their dependencies are not bundled and require separate installation and configuration; DonSeTch is explicit-only by default.
- Users upgrading from the former Hound integration must migrate because Hound was removed in version 4.0 in favor of the DonSeTch path.
How does this agent compare with similar options?
Compared with Hermes' native web_search and web_extract, Web Search Plus adds a multi-provider engine, fallback, caching, and quality reporting. Its optional wsp backend can route those native calls through the WSP engine, but merely installing the plugin does not select that backend. The repository also identifies web-search-plus-plugin as the TypeScript/OpenClaw edition, whereas this repository is the Python port for the Hermes Agent plugin API. DonSeTch is not a bundled substitute: it is an independent AGPL-3.0-only local MCP program connected through a separate stdio process.
Key facts side by side with the most closely related agents.
| Agent | Source review | Form / cost | Stars | Updated | Language | Full support on |
|---|---|---|---|---|---|---|
| Web Search Plus for Hermes This agent | 78 · Good | Agent plugin / skillFree | ★ 445 | 10d ago | Python | — |
| BambooAI Data Analyst | 70 · Some gaps | Web appFree + model costs | ★ 848 | 1d ago | Python | OpenAI API · Claude API |
| MiroThinker | 45 · Major gaps | CLIFree + model costs | ★ 8.5k | 8mo ago | Python | OpenAI API · Claude API |
| Tongyi DeepResearch | 44 · Major gaps | CLIFree + model costs | ★ 20k | 8mo ago | Python | — |
How does FollowAgents rate this agent?
Why each dimension lost points
The README, FAQ, and DonSeTch guide path explain that queries and URLs go to the selected external service or local sidecar, warn that provider privacy terms differ, and disclose local cache and health-state writes. This supports high data-flow transparency and source attribution. Keys are managed through the active Hermes profile and can be cleared, but the packet does not show key encryption, cache deletion, or per-query confirmation for sensitive inputs. Research mode can call multiple APIs, with budget controls but no per-call confirmation. Dependency security has a vulnerability-reporting path and CI installation details, but no evidence of dependency locking or auditing. Setup and disable options provide partial recovery.
The packet describes provider fallback, cooldowns, partial-result retention, one response per requested URL, and structured errors, alongside specific fix notes. CI configuration covers several Python versions, tests, and generated-file checks, but there is no execution evidence, so this review does not claim they passed. External services, quotas, rate limits, and the local sidecar remain availability dependencies. The README's new-release summary and FAQ's routing-order description differ slightly, reducing self-consistency.
Documentation covers Hermes users, CLI and gateway operation, commercial APIs, self-hosted SearXNG, explicit providers, and local DonSeTch, while distinguishing search from extraction. Explicit calls, auto_allow, provider boundaries, and fixed TinyFish/Octen endpoints are described concretely. Automatic intent routing still requires consulting the routing reference, and the product is limited to Hermes and Python 3.10+ environments.
The README organizes links for installation, configuration, operation, and deeper contracts. Quick installation, key setup, restart, upgrade, FAQ, examples, and limitations are well covered. The MIT license, current release, release-by-release changes, and security reporting route are clear. Naming varies somewhat among Web Search Plus, WSP, and repository labels such as schema tests. Maintenance responsibility is represented by a copyright name and repository security process, without an organizational support path or stated service period.
The two tools provide source search and page extraction, with documentation describing original links and optional provider, skip-reason, intent, and timing details. Multi-provider fallback and local options add practical value. Quotas, latency, privacy terms, and extra API calls in research mode impose costs; controls are documented, but there is no general cost guarantee.
The README links key features to the user guide, FAQ, changelog, provider reference, and routing reference. The CI workflow and test fixtures offer static evidence about implementation and test boundaries, while release notes list specific fixes. Performance and quality comparisons are mostly claims in the README; the packet does not include their datasets or independent corroboration. Static review cannot establish actual result correctness.
- Queries and target URLs are sent to the selected service; local or key-free does not mean offline or anonymous. Check the specific provider terms before sending sensitive content.
- The packet shows no dependency lock or audit evidence and includes no execution record; CI configuration and documented performance comparisons were not verified in this review.
- Research mode can increase API calls and cost; providers control quotas, pricing, and availability.
Recent releases
Latest entries from the project's GitHub Releases. Notes are quoted from upstream as written, not rewritten or tested by FollowAgents.
-
v5.1.9 v5.1.9 settings mean what they say
Release notes excerpt
Web Search Plus 5.1.9 Settings mean what they say. A provider you allow for automatic search is used, status names a placeholder key and says when nothing runs automatically, and a self-hosted SearXNG URL set the Hermes way is picked up. Found by an audit of the MCP server against this plugin and decided with the maintainer; checked with the real Hermes v0.22.5 CLI in a throwaway home. No configuration changes are needed. Fixed …View on GitHub ↗ -
v5.1.8 v5.1.8 · setup tells the truth
Release notes excerpt
Web Search Plus 5.1.8 Setup tells the truth. Whether a provider can run is now decided in one place, and every surface — status, setup, the tools' availability, the session hint, the engine and the native backend — shows that one answer, with the reason and the command that fixes it. Found by an analysis of the plugin's implicit state machines; checked with the real Hermes v0.22.5 CLI in a throwaway home. No configuration changes are needed. Fixed …View on GitHub ↗ -
v5.1.7 v5.1.7 · extract failures name their reason
Release notes excerpt
Web Search Plus 5.1.7 Honest page extraction. When a page cannot be read, the model now learns why and can tell the user, and Web Search Plus stops paying for providers that cannot help. Found in a code review and checked with real keys through the registered Hermes tool. No configuration changes are needed. Fixed …View on GitHub ↗
FAQ
Do I have to buy a search API?
How many providers must I configure?
web_search_plus, and one extraction-capable key or endpoint enables web_extract_plus. Additional providers mainly expand routing and fallback choices.Is DonSeTch installed with the plugin?
DONSETCH_BIN must point to its absolute executable path. The README identifies 4.2.9 as the adapter's tested version.What happens when a provider fails?
Can sensitive queries be sent to every provider safely?
Is Web Search Plus for Hermes free?
Is Web Search Plus for Hermes open source?
Is Web Search Plus for Hermes safe to use?
Is Web Search Plus for Hermes still maintained?
Who makes Web Search Plus for Hermes?
What is Web Search Plus for Hermes's official website?
Thanks — we use this to improve the page.