Persistent agent memory and identity via a single bsp() function operating on structured pscale blocks.
Do not connect
No critical issue was found, but the combined weight of the other findings put the score below 25/100. Remediate the findings before connecting this server.
Scanned 2 months ago Due for re-check
A server can change after it's graded. Re-run the automated scan to refresh this report.
This grade is deterministic and reproducible: the same server surface always yields the same grade under a given algorithm version. It is a real automated assessment computed by the MCPGrade engine from what the probe actually observed — not a fabricated or opinion score. It is not a manual human pentest, so it can miss context-specific risks.
Every signal below was measured directly by the automated probe. The grade is derived only from evidence like this — nothing is assumed.
The bsp-mcp-server MCP server exposes 12 tools, focused primarily on network capabilities. Its published description reads: "Persistent agent memory and identity via a single bsp() function operating on structured pscale blocks". It communicates over Streamable HTTP using the 2025-06-18 protocol revision, and does not require authorization to connect. MCPGrade currently rates bsp-mcp-server F — no critical issue was found, but the combined weight of the other findings put the score below 25/100, so the findings should be remediated before connecting. Its most notable findings include "Data-exfiltration parameters" and "Unguarded destructive tools". This report is a deterministic, reproducible automated assessment: the same observed surface always yields the same grade under a given algorithm version, and it is refreshed as new scans arrive and free to read — but it is not a substitute for a manual human security review. Always re-check bsp-mcp-server's advertised tools, transport security, and authorization posture before connecting an autonomous agent or sharing sensitive context with it.
Add bsp-mcp-server to an MCP client using the endpoint below. Review the grade and findings above before granting an autonomous agent access.
MCP endpoint
https://bsp.hermitcrab.me/mcp/v1Streamable HTTP transport. This is the MCP endpoint, not a website — paste it into your MCP client server list rather than a browser.
Captured passively during the read-only scan. Click any value to find servers that match it.
The tools this server advertises via tools/list — names, purposes, and the parameters each accepts, exactly as enumerated read-only.
The unified bsp() function. Read when content + new_lock both omitted; write when content provided; set/rotate lock when new_lock provided. Two coordinates: spindle (S, the address) and pscale_attenti
Sorted worst-first. Each finding shows its severity, what it means, its OWASP MCP Top-10 mapping, and a recommended fix — the check id links to the exact methodology row that produced it.
tool: bsp
A tool exposes a free-text/webhook/feedback parameter with no honest purpose — a classic exfil channel.
Lock semantics: secret = proof of current authority; new_lock = targetRecommendationRemove side-channel parameters; constrain tool inputs.
tool: bsp
Tools that execute commands / write files / make arbitrary requests are exposed without annotations or guards.
eft\" + block \"shell\" lands at the default beach as block \"shell:wRecommendationAnnotate destructive tools and require human approval.
tool: bsp
A tool description tries to alter the model’s use of another tool.
RecommendationDescriptions must describe only their own tool.
tool: bsp
One tool reads private data, ingests untrusted content, AND can send data outbound — the three preconditions for autonomous data theft in a single call. A prompt injection reaching this tool can exfiltrate secrets with no further step.
bsp() function. Read when content + new_lock both omitted; write when content proRecommendationSplit the capabilities across separate least-privilege tools so no single tool can read secrets, read untrusted input, and reach the network at once.
tool: bsp
The input schema declares a parameter that asks the caller to hand over a password, key, token, or other secret — a credential-harvest / token-passthrough risk. (Bare pagination `token` params are excluded.)
parameter "secret"RecommendationNever pass secrets as tool arguments; authenticate out-of-band and remove credential parameters from the schema.
tool: bsp-floor
Tools that execute commands / write files / make arbitrary requests are exposed without annotations or guards.
itions (a whole shell, or every block at a beach). The calling LLM isRecommendationAnnotate destructive tools and require human approval.
tool: bsp-floor
A tool description tries to alter the model’s use of another tool.
ry companion to bsp(). Lays two or more blocks against the common fRecommendationDescriptions must describe only their own tool.
tool: pscale_settle
Tools that execute commands / write files / make arbitrary requests are exposed without annotations or guards.
your sovereign shell (optional). Stored at the hidden directory of yRecommendationAnnotate destructive tools and require human approval.
tool: pscale_settle
A tool description tries to alter the model’s use of another tool.
later edits via bsp() as `secret`). Defaults to https://beach.happyRecommendationDescriptions must describe only their own tool.
tool: pscale_grain_reach
A tool description tries to alter the model’s use of another tool.
ing identity in bsp(). Defaults to https://beach.happyseaurchin.comRecommendationDescriptions must describe only their own tool.
tool: pscale_key_publish
The input schema declares a parameter that asks the caller to hand over a password, key, token, or other secret — a credential-harvest / token-passthrough risk. (Bare pagination `token` params are excluded.)
parameter "secret"RecommendationNever pass secrets as tool arguments; authenticate out-of-band and remove credential parameters from the schema.
tool: pscale_pool_engage
A tool exposes a free-text/webhook/feedback parameter with no honest purpose — a classic exfil channel.
ing per-subject history:<handle> spines) is the resolver's bsp() job, nRecommendationRemove side-channel parameters; constrain tool inputs.
tool: pscale_pool_engage
A tool description tries to alter the model’s use of another tool.
RecommendationDescriptions must describe only their own tool.
tool: pscale_pool_engage
The input schema declares a parameter that asks the caller to hand over a password, key, token, or other secret — a credential-harvest / token-passthrough risk. (Bare pagination `token` params are excluded.)
parameter "secret"RecommendationNever pass secrets as tool arguments; authenticate out-of-band and remove credential parameters from the schema.
tool: pscale_stream_engage
A tool description tries to alter the model’s use of another tool.
. Distinct from pscale_pool_engage, which owns a spool and a liquid buffer and serRecommendationDescriptions must describe only their own tool.
tool: pscale_stream_engage
The input schema declares a parameter that asks the caller to hand over a password, key, token, or other secret — a credential-harvest / token-passthrough risk. (Bare pagination `token` params are excluded.)
parameter "secret"RecommendationNever pass secrets as tool arguments; authenticate out-of-band and remove credential parameters from the schema.
tool: pscale_networking
A tool exposes a free-text/webhook/feedback parameter with no honest purpose — a classic exfil channel.
-chosen act."},"secret":{"type":"string","description":"Write authoritRecommendationRemove side-channel parameters; constrain tool inputs.
tool: pscale_networking
A tool description tries to alter the model’s use of another tool.
credit / SQ via pscale_verify_rider), and either PERCEIVE (default, permission='askRecommendationDescriptions must describe only their own tool.
tool: pscale_networking
The input schema declares a parameter that asks the caller to hand over a password, key, token, or other secret — a credential-harvest / token-passthrough risk. (Bare pagination `token` params are excluded.)
parameter "secret"RecommendationNever pass secrets as tool arguments; authenticate out-of-band and remove credential parameters from the schema.
tool: pscale_play
Tools that execute commands / write files / make arbitrary requests are exposed without annotations or guards.
t/history/stash/shell exist for the handle — the legacy names witnessRecommendationAnnotate destructive tools and require human approval.
tool: pscale_play
A tool description tries to alter the model’s use of another tool.
rld. Sibling of pscale_invite: invite is the welcome passage for a newcomer;RecommendationDescriptions must describe only their own tool.
tool: pscale_play
The input schema declares a parameter that asks the caller to hand over a password, key, token, or other secret — a credential-harvest / token-passthrough risk. (Bare pagination `token` params are excluded.)
parameter "secret"RecommendationNever pass secrets as tool arguments; authenticate out-of-band and remove credential parameters from the schema.
tool: pscale_genus
A tool exposes a free-text/webhook/feedback parameter with no honest purpose — a classic exfil channel.
own fold, note→history kernel-timestamped, refusals reported into condRecommendationRemove side-channel parameters; constrain tool inputs.
tool: pscale_genus
Tools that execute commands / write files / make arbitrary requests are exposed without annotations or guards.
f — the agent's shell as one nested whole, koan and clouds riding inRecommendationAnnotate destructive tools and require human approval.
tool: pscale_genus
A tool description tries to alter the model’s use of another tool.
hatch (fourteen bsp writes from any door); the first of the genus iRecommendationDescriptions must describe only their own tool.
tool: (server instructions)
A tool exposes a free-text/webhook/feedback parameter with no honest purpose — a classic exfil channel.
lies). A handle+secret is an EDIT-LATCH on a public page the operatorRecommendationRemove side-channel parameters; constrain tool inputs.
tool: (server instructions)
Tools that execute commands / write files / make arbitrary requests are exposed without annotations or guards.
n address. Full shell at bsp(agent_id="pscale", block="soft-agent", sRecommendationAnnotate destructive tools and require human approval.
tool: (server instructions)
A tool description tries to alter the model’s use of another tool.
bsp-mcp-server — two functions (bsp and the n-ary bRecommendationDescriptions must describe only their own tool.
The server accepts tool enumeration (and likely invocation) with no authentication.
RecommendationRequire OAuth 2.1 authorization for any server exposing non-public tools.
tool: bsp
The server advertises open-world / broadly-scoped capabilities.
annotations.destructiveHint = trueRecommendationScope tools to the minimum needed.
tool: bsp-floor
The server advertises open-world / broadly-scoped capabilities.
annotations.openWorldHint = trueRecommendationScope tools to the minimum needed.
tool: pscale_settle
The server advertises open-world / broadly-scoped capabilities.
annotations.openWorldHint = trueRecommendationScope tools to the minimum needed.
tool: pscale_settle
The description pairs a fetch imperative with a hardcoded external URL, letting the server relocate its real instructions off-metadata and mutate them after review. Fires only when a fetch verb and a URL co-occur, so benign documentation links do not.
`). Defaults to https://beach.happyseaurchin.com; pass agent_id for another beach.RecommendationDo not direct the model to fetch and act on external URLs; treat linked content as untrusted.
tool: pscale_grain_reach
The server advertises open-world / broadly-scoped capabilities.
annotations.openWorldHint = trueRecommendationScope tools to the minimum needed.
tool: pscale_key_publish
The server advertises open-world / broadly-scoped capabilities.
annotations.destructiveHint = trueRecommendationScope tools to the minimum needed.
tool: pscale_pool_engage
The server advertises open-world / broadly-scoped capabilities.
annotations.destructiveHint = trueRecommendationScope tools to the minimum needed.
tool: pscale_stream_engage
The server advertises open-world / broadly-scoped capabilities.
annotations.destructiveHint = trueRecommendationScope tools to the minimum needed.
tool: pscale_verify_rider
The server advertises open-world / broadly-scoped capabilities.
annotations.openWorldHint = trueRecommendationScope tools to the minimum needed.
tool: pscale_networking
The server advertises open-world / broadly-scoped capabilities.
annotations.destructiveHint = trueRecommendationScope tools to the minimum needed.
tool: pscale_play
The server advertises open-world / broadly-scoped capabilities.
annotations.openWorldHint = trueRecommendationScope tools to the minimum needed.
tool: pscale_genus
The server advertises open-world / broadly-scoped capabilities.
annotations.destructiveHint = trueRecommendationScope tools to the minimum needed.
tool: (server instructions)
The description pairs a fetch imperative with a hardcoded external URL, letting the server relocate its real instructions off-metadata and mutate them after review. Fires only when a fetch verb and a URL co-occur, so benign documentation links do not.
atch. - URL ("https://example.com") → that federated beach at <origin>/.weRecommendationDo not direct the model to fetch and act on external URLs; treat linked content as untrusted.
tool: pscale_pool_engage
An unusually long description is a common injection-padding tactic.
description length 2009 charsRecommendationKeep descriptions concise.
Vantaj uptime monitoring via MCP — manage monitors, heartbeats, incidents, and status pages.
Unified gateway to Algeria's TKAWEN ecosystem: commerce, certification, and AI tools.
Provides access to the Cohereon Doctrine AI safety framework with governance components, tiered access, and agent onboarding.
Agentic rails for complex workflows with receipts, fees, and MCP tool access.
Structural TC39 spec lookup for ECMA-262 and ECMA-402 in AI agents, SHA-pinned and offline-first.
Structural TC39 spec lookup for ECMA-262 and ECMA-402 in AI agents, SHA-pinned and offline-first.
The n-ary companion to bsp(). Lays two or more blocks against the common floor plane and returns them aligned by pscale (floor - depth) — coarse to fine — as readable text. The law: cross-block corres
Settle into a sedimentary collective — a public group on a federated beach where each new member lands at the next open position in arrival order (a sediment layer settling into rock). The beach assig
Establish a grain at a federated beach — first durable bilateral commitment. Symmetric: same call from either side. The beach detects state — first call creates the block and writes one side; second c
Derive an X25519+Ed25519 keypair from your secret + handle (Argon2id). Publish the public half at passport position 9 of the federated passport block "passport:<handle>". Private half is never stored.
TWO WRITE VERBS, chosen by where the text lands (BOTH are live — there is no single-verb 'submit-only' mode): contribution = APPEND to the pool (the shared spool everyone pulls; this is the basic pool
ONE WRITE VERB — `say` — because a stream has no buffer to stage into: your reading lands in YOUR OWN mirror (<field>:<handle>) at the attended address, and a mirror is revisable by its holder forever
Deterministic arithmetic check on a Level 2 ecosquared rider. Verifies: chain integrity (sha256 chain), credit conservation (rider.credits.n <= passport.6.1 balance), SQ recompute (Σ v_latest/giver_to
The SAND (Level 3) driver — the social neuron. Walk a committed channel (a grain, a pool, an accumulator like marks) for new rider-bearing probes since your marker, verify each deterministically (chai
The welcome — call this FIRST when a person arrives or asks what this place is. A bare call returns a Character-voiced director's note for the OPENING TURN: read who is actually about, surface the bea
Inhabit a handle in a world, in one call — the no-fiddle entry that makes 'play anya on thornwood' just work. Resolves the world to its beach (a sub-domain <world>.beach.<host>, or a full URL), engage
Wear a genus-one agent's mind for a wake — one call returns the instance's COMPOSED context window, byte-identical to what the kernel hands a bare-API LLM: SYSTEM (the recipe, the dehydrated index, th