Loading ToolBench
Generate random IDs, QR codes, and hashes, encode and decode values, and geolocate IPs, plus gated network and system diagnostics, via MCP. STDIO or Streamable HTTP.
Strong, consistent quality across 7 tools. All tools have clear names starting with action verbs (toolkit_*), comprehensive descriptions (172 - 370 chars), and complete JSON Schema with type definitions and parameter constraints. Output schemas are documented in descriptions. Tool names are distinct and specific (generate_id vs generate_qr, hash_value vs encode_value). No secrets in parameters. Error handling guidance present in descriptions (e.g., 'Decoding a value that is malformed... is reported as a recoverable error'). Minor gaps: output field details could be more formally structured as separate schema documentation; no per-tool permission annotations; error categorization (retryable vs. fatal) not explicit in all tools.
Run a read-only network diagnostic from the server host. mode selects the probe: ping (ICMP round-trip), traceroute (the hop path to the target), connectivity (a raw TCP connect to target on port), or public_ip (the host's own egress IP, where target is absent and ignored). target takes an IPv4/IPv6 address or a hostname (resolved server-side) and is required for every mode except public_ip; port is required only for connectivity and is passed separately, never as host:port. count sets ping echo requests and timeoutMs the per-probe deadline. A host that does not respond is reported as reachable:false — a valid result, not an error. The populated output fields depend on mode. This tool diagnoses the SERVER's own network, so it is useful only on a local/self-hosted deployment; reaching a private/reserved/internal target additionally requires the operator to enable TOOLKIT_ALLOW_PRIVATE_NETWORK.
Report a facet of the server host's system state, read-only. what selects the facet: os (platform, release, architecture, hostname, uptime, Node version), cpu (model, core count, speed), memory (total/free/used bytes), load (1/5/15-minute load averages, all zero on Windows), or interfaces (non-internal network interfaces with their addresses and families). Exactly one facet object is populated per call, matching what; the others are absent. All values describe the host this server runs on, NOT the calling client — so this is useful only on a local or self-hosted deployment. interfaces and os disclose host topology and version details, which is why this tool is gated off by default.
Encode or decode a value across base64, base64url, hex, or URL (percent) encoding, in either direction. Set operation to "encode" to transform raw UTF-8 text into the chosen encoding, or "decode" to recover the original text from an encoded value. base64url uses the URL-safe alphabet (- and _ instead of + and /); url applies encodeURIComponent / decodeURIComponent. Decoding a value that is malformed for the chosen encoding is reported as a recoverable error, not a silent best-effort.
Output schemas are documented only in prose within descriptions, not as formal JSON Schema. For toolkit_check_system, toolkit_check_network, and toolkit_geolocate_ip, the exact field structure and types of return values are implicit rather than explicitly registered.
No per-tool permission annotations. Tools like toolkit_check_system and toolkit_check_network disclose system topology and version details (marked 'gated off by default'), but there is no formal scope declaration (e.g., 'requires:read:system') in the tool metadata to enable least-privilege agent configuration.
toolkit_geolocate_ip description states results are 'best-effort' and fields can be absent, but does not provide explicit guidance on how an agent should handle missing fields (e.g., 'If latitude/longitude is absent, the location is unreliable').
Inferred effective spec: 2026-07-28+.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-22 | A | 88 | 2026-07-28+ | v2 |
| 2026-03-09 | D | 50 | - | v1 |
Mint cryptographically-random identifiers using the platform CSPRNG — the correct source for IDs that must be unpredictable, unlike model-generated values. type selects the format: uuid_v4 (random, the default), uuid_v7 (time-ordered, sortable by creation), or ulid (26-char Crockford-base32, lexicographically sortable). Set count to mint a batch in one call (up to 1000); the returned ids array always contains exactly count values and is never truncated. For uuid_v7 and ulid, a batch is monotonic — strictly increasing even within the same millisecond — so the ids array stays in sorted creation order. IDs from this tool feed into toolkit_generate_qr (pass ids[0] as data) to create a scannable code.
Encode text or a URL into a QR code. data is the content to encode (a link, a generated identifier such as toolkit_generate_id's ids[0], or any string). format selects the output: svg returns inline SVG markup, png_base64 returns base64-encoded PNG bytes (with mimeType and byteLength), and terminal returns a block of Unicode block characters renderable in a monospace terminal. errorCorrection (L/M/Q/H) trades data capacity for damage tolerance, margin sets the quiet-zone width, and scale sets pixels per module for raster output. The returned version (1–40) reflects how dense the encoded data is. png_base64 renders (modules + 2 × margin) × scale pixels per side and rejects anything past 2048 px with a typed raster_too_large error, so a dense symbol needs a lower scale; svg carries no such limit.
Resolve a public IP address (or hostname) to geographic and network metadata: country, region, city, latitude/longitude, the owning ASN and organization, timezone, and the proxy/hosting/mobile quality flags. target accepts an IPv4/IPv6 address or a hostname — a hostname is DNS-resolved first and the resolvedIp field echoes which IP was actually located. The provider is called directly (never the target), so this is SSRF-free and safe to expose anywhere. Results are best-effort and provider-bounded: VPNs, proxies, mobile NAT, and anycast all defeat IP-to-location, accuracy is city-level at best, and many fields can be absent for reserved or thinly-documented ranges — absent fields are reported as unknown, never invented. Read proxy, hosting, and mobile before trusting the coordinates: a true on any of them means the location describes infrastructure, not the user. Private/reserved addresses have no public geolocation and are rejected. The source field names which provider answered.
Generate a cryptographic digest of a value, or verify a value against an expected digest. Set operation to "generate" for a lowercase-hex digest, or "compare" to constant-time-check value against the expected digest — compare is timing-safe and avoids manual string equality checks. Algorithm defaults to sha256; sha512 is also secure, while md5 and sha1 are exposed for checksum and file-integrity compatibility ONLY and must not be used for passwords, signatures, or any security purpose. inputEncoding controls how value and expected are read before hashing (utf8 default, or hex/base64 for raw binary data) so binary blobs need no decode round-trip. The canonical use is matching a download against a vendor-published checksum.
Error types and recovery paths are mentioned in prose but not formally categorized. For toolkit_generate_qr, 'data_too_large' and 'raster_too_large' are typed errors, but other tools lack explicit error classification to help agents decide whether to retry or escalate.