io.github.robinlidberg-dot/cap-shield is an MCP server published by io.github.robinlidberg-dot, hosted at cap-shield-robin.fly.dev. When last checked on 2026-11-22 it was working: it answers and lists its tools. It declares 9 tools, including measure_traffic, list_packages, remember, assemble_context, assemble_escalation_context.
Context selection and compression for AI agents - recall measured, not claimed.
Nothing else in the archive shares these values.
Squares are this server's values; dots are other servers sharing them. A shared address or publisher is a lead, not proof of a common operator.
Inferred from the host name and headers: what a server runs on, not who runs it.
Measure how much of your own agent traffic could be saved. NO ACCOUNT OR KEY NEEDED — use this first. Returns byte savings over the wire and, if a query is given, token savings from selective context retrieval. Nothing is stored: the text is compressed in memory and discarded. Rate limited to 20 calls per hour per IP.
| Parameter | Type | Description |
|---|---|---|
| query | string | Optional. What your agent would search its memory for. Given this, token savings are measured too. |
| texts * | array | Your actual messages. 200 max. |
| dict_id | string | Which trained dictionary to measure against. Call list_packages to see the options and their measured figures. |
{
"name": "measure_traffic",
"title": "Measure possible savings",
"description": "Measure how much of your own agent traffic could be saved. NO ACCOUNT OR KEY NEEDED — use this first. Returns byte savings over the wire and, if a query is given, token savings from selective context retrieval. Nothing is stored: the text is compressed in memory and discarded. Rate limited to 20 calls per hour per IP.",
"inputSchema": {
"type": "object",
"required": [
"texts"
],
"properties": {
"query": {
"type": "string",
"description": "Optional. What your agent would search its memory for. Given this, token savings are measured too."
},
"texts": {
"type": "array",
"items": {
"type": "string"
},
"description": "Your actual messages. 200 max."
},
"dict_id": {
"type": "string",
"default": "finance",
"description": "Which trained dictionary to measure against. Call list_packages to see the options and their measured figures."
}
}
},
"annotations": {
"title": "Measure possible savings",
"readOnlyHint": true,
"openWorldHint": false,
"idempotentHint": true,
"destructiveHint": false
}
}List the available dictionaries with their MEASURED compression, including the ones that perform badly. Each entry says whether it works one message at a time or only batched, and how many messages came out LARGER. No key needed.
No parameters.
{
"name": "list_packages",
"title": "List compression packages",
"description": "List the available dictionaries with their MEASURED compression, including the ones that perform badly. Each entry says whether it works one message at a time or only batched, and how many messages came out LARGER. No key needed.",
"inputSchema": {
"type": "object",
"properties": {}
},
"annotations": {
"title": "List compression packages",
"readOnlyHint": true,
"openWorldHint": false,
"idempotentHint": true,
"destructiveHint": false
}
}Store a memory entry for later retrieval. REQUIRES A KEY. This does not call any language model — it stores text in an isolated per-tenant archive. Use assemble_context to get relevant entries back.
| Parameter | Type | Description |
|---|---|---|
| text * | string | |
| item_id | string | Optional. Same id overwrites the entry. Omit for a generated one. |
| ttl_days | integer | Optional. After this many days, assemble_context stops selecting this entry. It is not deleted, just no longer chosen. Omit to keep it eligible indefinitely, as before this field existed. |
| namespace * | string | Groups related memories. |
{
"name": "remember",
"title": "Store a memory entry",
"description": "Store a memory entry for later retrieval. REQUIRES A KEY. This does not call any language model — it stores text in an isolated per-tenant archive. Use assemble_context to get relevant entries back.",
"inputSchema": {
"type": "object",
"required": [
"namespace",
"text"
],
"properties": {
"text": {
"type": "string"
},
"item_id": {
"type": "string",
"description": "Optional. Same id overwrites the entry. Omit for a generated one."
},
"ttl_days": {
"type": "integer",
"description": "Optional. After this many days, assemble_context stops selecting this entry. It is not deleted, just no longer chosen. Omit to keep it eligible indefinitely, as before this field existed."
},
"namespace": {
"type": "string",
"description": "Groups related memories."
}
}
},
"annotations": {
"title": "Store a memory entry",
"readOnlyHint": false,
"openWorldHint": false,
"idempotentHint": false,
"destructiveHint": false
}
}Retrieve the memory entries that answer a question, within a token budget. REQUIRES A KEY. Send the returned 'context' to your language model INSTEAD of the whole history. This does not call a model itself — it selects what to send. The budget is a ceiling, not a target: selection stops where relevance runs out, often well below it. The response says how many entries were left behind and why. If the response has a 'budget_warning', the best match did not fit: call again with its suggested_token_budget.
| Parameter | Type | Description |
|---|---|---|
| query * | string | What you need context about. |
| reduce | boolean | Opt-in (FAS L): shorten the assembled context with a small, local, task-agnostic model before it is returned, to lower your own model's token cost. Requires tenant consent for context_reduction (set via /api/v1/gdpr/consent) — otherwise the call is rejected with SAMTYCKE_KRAVS instead of silently ig… |
| namespace * | string | |
| token_budget | integer | |
| target_tokens | integer | Only relevant together with reduce=true. How many tokens the assembled context should be shortened to. Omit for a default halving of what was actually retrieved. |
{
"name": "assemble_context",
"title": "Retrieve relevant memory",
"description": "Retrieve the memory entries that answer a question, within a token budget. REQUIRES A KEY. Send the returned 'context' to your language model INSTEAD of the whole history. This does not call a model itself — it selects what to send. The budget is a ceiling, not a target: selection stops where relevance runs out, often well below it. The response says how many entries were left behind and why. If the response has a 'budget_warning', the best match did not fit: call again with its suggested_token_budget.",
"inputSchema": {
"type": "object",
"required": [
"namespace",
"query"
],
"properties": {
"query": {
"type": "string",
"description": "What you need context about."
},
"reduce": {
"type": "boolean",
"default": false,
"description": "Opt-in (FAS L): shorten the assembled context with a small, local, task-agnostic model before it is returned, to lower your own model's token cost. Requires tenant consent for context_reduction (set via /api/v1/gdpr/consent) — otherwise the call is rejected with SAMTYCKE_KRAVS instead of silently ignoring the request. NOT query-aware: selection is driven by what is least informative in general, not by this query."
},
"namespace": {
"type": "string"
},
"token_budget": {
"type": "integer",
"default": 2000
},
"target_tokens": {
"type": "integer",
"description": "Only relevant together with reduce=true. How many tokens the assembled context should be shortened to. Omit for a default halving of what was actually retrieved."
}
}
},
"annotations": {
"title": "Retrieve relevant memory",
"readOnlyHint": true,
"openWorldHint": false,
"idempotentHint": true,
"destructiveHint": false
}
}Retrieve a WIDE context for a question your model could not answer from assemble_context. REQUIRES A KEY. Use it when assemble_context returned escalate_recommended=true, or when the answer depends on dates, counts or several sessions. This does not call a model itself. Returns 'context' (up to token_budget, default 150 000 tokens: lower it if your model's window is smaller), where each stored entry starts with a header '--- Session YYYY/MM/DD (Ddd) HH:MM ---' giving when it was stored, and 'utelamnat_kontext': lower-ranked entries that were left out, in the same format. Read those only if the answer is still missing from 'context'. Python users can pass the whole response to cap_escalate_helper.eskalera_om_behovs_v6().
| Parameter | Type | Description |
|---|---|---|
| query * | string | The question to answer. |
| namespace * | string | |
| token_budget | integer | Ceiling for 'context'. The default is what the method was measured with. |
| utelamnat_tokens | integer | Ceiling for 'utelamnat_kontext'. 0 = none. |
{
"name": "assemble_escalation_context",
"title": "Retrieve wide dated memory",
"description": "Retrieve a WIDE context for a question your model could not answer from assemble_context. REQUIRES A KEY. Use it when assemble_context returned escalate_recommended=true, or when the answer depends on dates, counts or several sessions. This does not call a model itself. Returns 'context' (up to token_budget, default 150 000 tokens: lower it if your model's window is smaller), where each stored entry starts with a header '--- Session YYYY/MM/DD (Ddd) HH:MM ---' giving when it was stored, and 'utelamnat_kontext': lower-ranked entries that were left out, in the same format. Read those only if the answer is still missing from 'context'. Python users can pass the whole response to cap_escalate_helper.eskalera_om_behovs_v6().",
"inputSchema": {
"type": "object",
"required": [
"namespace",
"query"
],
"properties": {
"query": {
"type": "string",
"description": "The question to answer."
},
"namespace": {
"type": "string"
},
"token_budget": {
"type": "integer",
"default": 150000,
"description": "Ceiling for 'context'. The default is what the method was measured with."
},
"utelamnat_tokens": {
"type": "integer",
"default": 60000,
"description": "Ceiling for 'utelamnat_kontext'. 0 = none."
}
}
},
"annotations": {
"title": "Retrieve wide dated memory",
"readOnlyHint": true,
"openWorldHint": false,
"idempotentHint": true,
"destructiveHint": false
}
}Get an account and an API key. Requires an email address. The key is returned ONCE and cannot be shown again — store it immediately. Beta quotas are low by design; they are hard stops, never overage billing.
| Parameter | Type | Description |
|---|---|---|
| email * | string | |
| traffic | string | Optional. What your agent sends. |
{
"name": "get_account",
"title": "Create account and API key",
"description": "Get an account and an API key. Requires an email address. The key is returned ONCE and cannot be shown again — store it immediately. Beta quotas are low by design; they are hard stops, never overage billing.",
"inputSchema": {
"type": "object",
"required": [
"email"
],
"properties": {
"email": {
"type": "string"
},
"traffic": {
"type": "string",
"description": "Optional. What your agent sends."
}
}
},
"annotations": {
"title": "Create account and API key",
"readOnlyHint": false,
"openWorldHint": false,
"idempotentHint": false,
"destructiveHint": false
}
}Issue a short-lived (15 minute) signed credential (a JWT) stating who the caller acts for. REQUIRES A KEY. Anyone can verify it independently against /.well-known/jwks.json — without calling CAP-Shield again. 'acts_for' and 'scope' are NOT validated against reality: the pass only attests that the key holder claimed this, at this time, and that the claim is recorded in the audit chain. This is identity and claimed authorization, not a payment or access-control mechanism.
| Parameter | Type | Description |
|---|---|---|
| scope | object | Optional, freeform, NOT validated. Example: {"can_read": "invoice_json", "max_budget_usd": 500}. |
| acts_for * | string | Who the agent claims to act for, e.g. 'acme-corp/procurement'. |
| audience | string | Optional. Who the pass is shown to. |
{
"name": "issue_pass",
"title": "Issue a signed agent pass",
"description": "Issue a short-lived (15 minute) signed credential (a JWT) stating who the caller acts for. REQUIRES A KEY. Anyone can verify it independently against /.well-known/jwks.json — without calling CAP-Shield again. 'acts_for' and 'scope' are NOT validated against reality: the pass only attests that the key holder claimed this, at this time, and that the claim is recorded in the audit chain. This is identity and claimed authorization, not a payment or access-control mechanism.",
"inputSchema": {
"type": "object",
"required": [
"acts_for"
],
"properties": {
"scope": {
"type": "object",
"description": "Optional, freeform, NOT validated. Example: {\"can_read\": \"invoice_json\", \"max_budget_usd\": 500}."
},
"acts_for": {
"type": "string",
"description": "Who the agent claims to act for, e.g. 'acme-corp/procurement'."
},
"audience": {
"type": "string",
"description": "Optional. Who the pass is shown to."
}
}
},
"annotations": {
"title": "Issue a signed agent pass",
"readOnlyHint": false,
"openWorldHint": false,
"idempotentHint": false,
"destructiveHint": false
}
}Pack several messages into ONE compressed packet instead of packing them one at a time. REQUIRES A KEY. Worth it for many small/medium messages (the per-message header cost is paid once per batch, and zstd shares context across them) — for a single message this has no advantage. Returns packet_b64 (hand it to unpack_batch to get the messages back) plus the measured byte savings versus packing each message separately. SECURITY: only batch messages that already share the SAME trust boundary (same tenant/agent — nothing attacker-influenced mixed in with a secret). Batched messages share one compression context, which can leak size-based information between them (a CRIME/BREACH-style side channel). There is no padding mitigation on this endpoint today, so trust-boundary separation is the only protection — don't rely on padding.
| Parameter | Type | Description |
|---|---|---|
| texts * | array | The messages to batch together, in order. Two or more — a single message gets no benefit. |
| dict_id | string | Which trained dictionary to compress against. Call list_packages to see the options. |
| agent_id | string | Optional. A label for audit/metrics. |
{
"name": "pack_batch",
"title": "Compress a batch of messages",
"description": "Pack several messages into ONE compressed packet instead of packing them one at a time. REQUIRES A KEY. Worth it for many small/medium messages (the per-message header cost is paid once per batch, and zstd shares context across them) — for a single message this has no advantage. Returns packet_b64 (hand it to unpack_batch to get the messages back) plus the measured byte savings versus packing each message separately. SECURITY: only batch messages that already share the SAME trust boundary (same tenant/agent — nothing attacker-influenced mixed in with a secret). Batched messages share one compression context, which can leak size-based information between them (a CRIME/BREACH-style side channel). There is no padding mitigation on this endpoint today, so trust-boundary separation is the only protection — don't rely on padding.",
"inputSchema": {
"type": "object",
"required": [
"texts"
],
"properties": {
"texts": {
"type": "array",
"items": {
"type": "string"
},
"description": "The messages to batch together, in order. Two or more — a single message gets no benefit."
},
"dict_id": {
"type": "string",
"default": "finance",
"description": "Which trained dictionary to compress against. Call list_packages to see the options."
},
"agent_id": {
"type": "string",
"description": "Optional. A label for audit/metrics."
}
}
},
"annotations": {
"title": "Compress a batch of messages",
"readOnlyHint": true,
"openWorldHint": false,
"idempotentHint": true,
"destructiveHint": false
}
}Unpack a batch packet made by pack_batch (or by the /api/v1/pack-batch endpoint) back into the exact original list of messages, in order. REQUIRES A KEY. Not for a packet made by the single-message pack endpoint — that is a different format and this will reject it.
| Parameter | Type | Description |
|---|---|---|
| agent_id | string | Optional. A label for audit/metrics. |
| packet_b64 * | string | The packet_b64 returned by pack_batch. |
{
"name": "unpack_batch",
"title": "Decompress a batch packet",
"description": "Unpack a batch packet made by pack_batch (or by the /api/v1/pack-batch endpoint) back into the exact original list of messages, in order. REQUIRES A KEY. Not for a packet made by the single-message pack endpoint — that is a different format and this will reject it.",
"inputSchema": {
"type": "object",
"required": [
"packet_b64"
],
"properties": {
"agent_id": {
"type": "string",
"description": "Optional. A label for audit/metrics."
},
"packet_b64": {
"type": "string",
"description": "The packet_b64 returned by pack_batch."
}
}
},
"annotations": {
"title": "Decompress a batch packet",
"readOnlyHint": true,
"openWorldHint": false,
"idempotentHint": true,
"destructiveHint": false
}
}Names, descriptions and schemas were written by the server and are shown as data.
Pattern matches on what the server declares. They show what it says it can do, not what it does.
None yet. Changes appear after a server's second daily check.