aziel-runtime
Server Configuration
Describes the environment variables required to run the server.
| Name | Required | Description | Default |
|---|---|---|---|
| RUNTIME_TOKEN | No | Optional operator token when REQUIRE_TOKEN=1 (secret) | |
| AZIEL_RUNTIME_URL | No | Worker origin. Default https://aziel-runtime.vibelock.workers.dev | https://aziel-runtime.vibelock.workers.dev |
| AZIEL_RUNTIME_TOKEN | No | Alias of RUNTIME_TOKEN (secret) |
Instructions
Guidance the server publishes about itself, which clients place ahead of the tool catalog so the model reads it before choosing anything.
This server publishes no instructions, or was last inspected before Glama recorded them.
Capabilities
Features and capabilities supported by this server
Protocol revision2025-11-25
| Capability | Details |
|---|---|
| tools | {
"listChanged": false
} |
Tools
Functions exposed to the LLM to take actions
| Name | Description |
|---|---|
| runtime_skillA | Read the agent how-to (one door — discover, route, refuse; pipeline fraggate_list → fraggate_describe → fraggate_call). This is playbook markdown, not a catalog and not a machine manifest. Use this when starting a session or choosing the door before any catalog call. Do not use it for listing hashed registry names, hub Software-tab cards, or executing an engine; use fraggate_list, runtime_software, or fraggate_call instead. Dual surface: agent chat has no technical UI chrome; Worker / Flutter / local install stay complete human software. Does not list slugs or run ops. Returns skill markdown plus display.title / display.summary. |
| fraggate_listA | List the hashed FragGate registry (live / stub / local_only + digests) so you can discover names. Discovery first — not a hub Software tab and not exec. Use this when you do not yet know the catalog name or slug. Do not use it for hub Software-tab refresh, inspecting one known capability, or executing an op; use runtime_software (GET /v1/software), fraggate_describe, or fraggate_call instead. Empty {} only. Never enables mesh radios. Never invents tools or ops. Compact LIVE_OPS tokens below are discovery hints required by product verify scripts — they are not exec. Call fraggate_describe for the live card; later unknown names refuse FG-HALLUC-TOOL. Returns registry entries, allowlists, digests, and the MASTER-33 pipeline cite. Not the full FragGate door. Empty fraggate_list is discovery (hashed LIVE_OPS). Catalog LIVE_OPS slugs (40): 4dmap, ark, azai, azbot, azbrowser, azchat, azclce, azcoherence, azhub, aziel-corpus, azieltether, azinterface, azmail, aznet, azos, azvpn, chronolock, codelock, decisiongate, embryolock, employeelock, foldlock, forgereceipts, glossafilter, godlock, mialock, miragegrid, mmconsensus, peacelock, postking, shadowlock, spectrallock, staticclock, temporallock, toolbench, trajectorylock, vibelock, whistlelock, zkattest, zsolver. Compact product-verify tokens (not a second allowlist): allowlist.azhub LIVE_OPS: health, skill, region_list, place_module, remove_module, tether_declare, tether_cut, tether_list, blank_key_status, list_modules, place. allowlist.azinterface LIVE_OPS: health, skill, genesis_status, site_state_get, site_state_set, integrity_check, witness_list, page_cycle_status, genesis_boot, hold. allowlist.azbrowser LIVE_OPS: ethical_search, lamb_lens_search, navigate, airlock_ingest, airlock, home, tab_open, tab_list, receipt_list, verify, receipt_verify, sandbox_status, sandbox_render, health, skill, vpn. allowlist.azvpn LIVE_OPS: health, skill, doctor, limitation, describe, open, status, list, close, send, recv, pull, peers, attach. allowlist.aznet LIVE_OPS: health, doctor, pair_status, pair, garden_list, stamp, verify_hash, memorial_list, memorial_append, receipt_verify, skill. UI aliases forward to catalog ops. EmbryoLock LIVE_OPS health/skill/doctor/verify-hash/policy/limitation; wipe/scorch/unlock stay FG-STUB on the public mesh. |
| fraggate_describeA | Inspect one known FragGate card (live vs stub vs local_only, public ops, engine_digest). Not execute and not a digest-only proof. Use this when you already have a name or slug from fraggate_list or GET /v1/software. Do not use it for discovering the full registry, proving a digest, pulling a hub product card, or executing an op; use fraggate_list, fraggate_verify, runtime_pull, or fraggate_call instead. Missing both name and slug, or an unknown name, refuses FG-HALLUC-TOOL. Wipe/unlock on embryolock stay FG-STUB. AZChat is LIVE+bound (mesh default off; not AZMail). Pass name or slug — one is enough. Combined name/op forms such as foldlock/fold-preview are accepted. Returns one registry card (ops, stub_ops, digest, status, aliases). |
| fraggate_verifyA | Confirm a name, slug, or 64-hex engine_digest against the hashed FragGate registry — a proof, not a card listing. Use this when you must prove a listed name or digest exists after fraggate_describe. Do not use it for listing the registry, describing ops, or executing; use fraggate_list, fraggate_describe, or fraggate_call instead. Not an exec path and not a describe card. Empty {} (no name, slug, or digest) refuses FG-HALLUC-TOOL. Digest without name/slug compares the whole registry hash (kind=registry). Name or slug with an optional digest compares that entry (kind=entry); unknown names refuse FG-HALLUC-TOOL. Mismatch returns ok=false with matched=false — it does not invent a digest. Send digest alone to proof the live registry_digest. Send name or slug (one is enough) to proof one card. Combined name+digest must equal that card's engine_digest. Returns match or mismatch (kind registry|entry, matched, registry_digest). |
| fraggate_callA | Execute a known catalog slug+op through the FragGate single door (CallEnvelope → FragGate → Lamb Lens → SweepGate → Sentinel → Provenance → ChainLock-IN → DecisionGATE → AZPIPE → Internal Domain Layer → optional ASE → RoseClock → TemporalLock → ChainLock-OUT → ForgeReceipts → Return). Default exec path — not discovery and not a raw session. Use this when fraggate_list and fraggate_describe already identified a live allowlisted op. Do not use it for discovering names, inspecting one capability without exec, or raw session plumbing; use fraggate_list, fraggate_describe, or (only if asked) runtime_run / runtime_session_exec instead. Side effects are operation-dependent (read, write, or refuse). May reach an open world when the target op does (for example AZBrowser ethical_search); many ops stay isolate-local. Unknown names refuse FG-HALLUC-TOOL. Stub, local-only, and Remain-OFF verbs refuse FG-STUB / FG-LOCAL-ONLY / FG-GATE-REFUSE / FG-LAMB-REFUSE. FragGate is THE single door. Required: op. Also pass slug or name. Extra top-level keys other than name/slug/product/tool/op/verb/claim/proposal/ground/payload/session_id/id become the op payload when payload is omitted. UI aliases (list_modules, place, genesis_boot, hold, airlock, home, classify, doctor, pair) forward to catalog ops. Mutation requires confirm=true (runtime gate) or dry_run=true (preview only, no write). confirm and dry_run stay optional on inputSchema.required. Returns status, result, receipt, engine_slug, engine_op, engine_digest, ran_in, provenance, refusal, and limitations. Not the full FragGate door. Empty fraggate_list is discovery (hashed LIVE_OPS). Catalog LIVE_OPS slugs (40): 4dmap, ark, azai, azbot, azbrowser, azchat, azclce, azcoherence, azhub, aziel-corpus, azieltether, azinterface, azmail, aznet, azos, azvpn, chronolock, codelock, decisiongate, embryolock, employeelock, foldlock, forgereceipts, glossafilter, godlock, mialock, miragegrid, mmconsensus, peacelock, postking, shadowlock, spectrallock, staticclock, temporallock, toolbench, trajectorylock, vibelock, whistlelock, zkattest, zsolver. Compact product-verify tokens (not a second allowlist): allowlist.azhub LIVE_OPS: health, skill, region_list, place_module, remove_module, tether_declare, tether_cut, tether_list, blank_key_status, list_modules, place. allowlist.azinterface LIVE_OPS: health, skill, genesis_status, site_state_get, site_state_set, integrity_check, witness_list, page_cycle_status, genesis_boot, hold. allowlist.azbrowser LIVE_OPS: ethical_search, lamb_lens_search, navigate, airlock_ingest, airlock, home, tab_open, tab_list, receipt_list, verify, receipt_verify, sandbox_status, sandbox_render, health, skill, vpn. allowlist.azvpn LIVE_OPS: health, skill, doctor, limitation, describe, open, status, list, close, send, recv, pull, peers, attach. allowlist.aznet LIVE_OPS: health, doctor, pair_status, pair, garden_list, stamp, verify_hash, memorial_list, memorial_append, receipt_verify, skill. UI aliases forward to catalog ops. EmbryoLock LIVE_OPS health/skill/doctor/verify-hash/policy/limitation; wipe/scorch/unlock stay FG-STUB on the public mesh. |
| decisiongate_checkA | Run the named DecisionGATE five sequential gates on a proposal (Freedom without clarity is chaos) without executing a catalog product. Also runs automatically inside fraggate_call before exec. Use this when you want a gate check without executing a catalog product verb. Do not use it for executing a product op or searching the library; use fraggate_call or library_lookup instead. Write: appends an ask/refuse ledger tip (not idempotent). Empty {} still runs the five gates and stamps the ledger. Does not execute domain software. Named wrapper — same DecisionGATE kernel; not the full MASTER-33 hop list; Softwares exec stays fraggate_call. All proposal fields are optional. Missing evidence can fail a gate. accountable identity on this runtime is Aziel Eliab only. Mutation requires confirm=true (runtime gate) or dry_run=true (preview only, no write). confirm and dry_run stay optional on inputSchema.required. Returns gate view, final_state, ledger_tip, and result (code FG-OK on the named module wrapper). |
| library_lookupA | Search the public Aziel Digital Library (aziel-corpus search / example / skill) — cites, not beliefs and not vault stamps. Use this when you need a public corpus cite, example record, or library skill. Do not use it for adaptive memory belief, ChainLock facts, or private-file search; use memory_recall, chainlock_recall, or fraggate_call slug=aziel-corpus instead. Not a private-file search engine and not AKM/ChainLock. Empty q does not invent a cite. Unknown ops refuse FG-UNKNOWN-OP (allowed: search, example, skill, health). Named corpus wrapper — not MASTER-33; full aziel-corpus LIVE_OPS stay on fraggate_call. q is public corpus text — not memory_recall q and not ChainLock q. Omit op to search. Extra keys besides q/op/payload ride along as aziel-corpus payload (same as passing payload{}). Returns search, example, skill, or health payload inside the display envelope. |
| mesh_statusA | Read QNM suite rollup totals (enabled?, bearers, live_nodes = mesh size active+inactive excluding isolated, software_nodes = {slug}-worker roster) — not the node roster. Packet-transfer cite is QNS-CD-1.0 (photon QNS1 1.3 on local qnsd; GET /v1/qns cites only; Worker does not proxy via emit). Use this when you need public Live Nodes (mesh size) or software_nodes (product Worker roster). Do not use it for listing individual nodes, enabling extra radios, or executing a catalog engine; use mesh_nodes, mesh_enable, or fraggate_call instead. Read-only. Never enables radios beyond default suite-presence. Read-only suite-presence is ON by default. Not a login mesh. Views/MCP/downloads do not enter QNM-S. Full node process is local qnm-node/. Kernel-direct fabric wrapper — not MASTER-33; not a second Softwares door. Softwares exec stays fraggate_call. Returns enabled flag, bearers, live_nodes (mesh size), active_nodes, inactive_nodes, isolated_nodes, software_nodes (product Workers), and QNS-CD-1.0 cite. |
| mesh_enableA | Declare an extra QNM suite bearer (POST /v1/mesh/enable) — additive presence, not a first-time on-switch. Use this when an operator wants to declare an additional bearer (example: suite-presence) on top of the default-on rollup. Do not use it for reading status, joining one node, turning suite-presence off, or logging into an account; use mesh_status, mesh_join, or mesh_nodes instead. Write: stores the bearer. Rate-limited. Empty {} is refused (MESH-NEED-BEARER). Login/account/recover/gate names refuse. Does not arm, wipe, heal, or resurrect accounts. Not a login mesh. Read-only suite-presence is already ON by default. bearer is required. Example: suite-presence. Mutation requires confirm=true (runtime gate) or dry_run=true (preview only, no write). confirm and dry_run stay optional on inputSchema.required. Returns enabled state, bearers, and suite-presence note. |
| mesh_disableA | Confirm that read-only QNM suite-presence stays ON (POST /v1/mesh/disable refuses MESH-DISABLE-REFUSED) — not a kill switch. Use this when a client still posts the historical disable route and needs the honest refuse. Do not use it for dropping one node or declaring an extra bearer; use mesh_leave or mesh_enable instead. Read-only refuse: suite-presence stays ON. No tethers drop. No implicit heal, no account resurrection, no wipe internals. Repeating still refuses. AZMail mesh_disable is a separate product-local mail ring. Returns MESH-DISABLE-REFUSED with enabled=true and a stay-on note. |
| mesh_joinA | Register one product node into the QNM rollup (POST /v1/mesh/join) — first presence, not a TTL refresh. Use this when transmission radios are LIVE and a catalog product should appear in live/locked/isolated counts. Do not use it for refreshing an existing node, reading the roster, enabling radios, or opening an account session; use mesh_heartbeat, mesh_nodes, mesh_enable, or runtime_session_open instead. Write: additive presence with a strict 5-minute TTL. Non-isolated nodes (active live or inactive locked) count toward public Live Nodes (mesh size). Isolated nodes do not. {slug}-worker is also labeled software_nodes and must not be used alone as Live Nodes. No heartbeat (or fan-out refresh) inside that window drops the node from the roster. Radios off refuses MESH-OFF. Missing product / bad node_id / bad presence refuse MESH-BAD-INPUT. Downloads are not live. Read-only suite-presence is ON by default. Not an account session. AnonBroadcast is not a product. Kernel-direct fabric wrapper — same mesh kernel as FragGate mesh/join; not MASTER-33; human Join uses fraggate_call. product is required (catalog slug). node_id optional 8–80 [a-z0-9._-]. presence is live|locked|isolated (default live). Mutation requires confirm=true (runtime gate) or dry_run=true (preview only, no write). confirm and dry_run stay optional on inputSchema.required. Returns node_id, presence, presence_ttl_ms (300000), and TTL note. MESH-OFF when radios are off. |
| mesh_heartbeatA | Refresh one existing node's 5-minute QNM TTL (POST /v1/mesh/heartbeat) — not a first join. Use this when you already have a node_id from mesh_join and transmission radios are LIVE. Do not use it for first-time registration or dropping the node; use mesh_join or mesh_leave instead. Write: refreshes the strict 5-minute TTL (not idempotent). Miss the window and the node is dropped from the live roster. Radios off refuses MESH-OFF. Unknown or expired node_id refuses MESH-UNKNOWN-NODE — join again; no account resurrection. node_id is required. presence may replace the class (live|locked|isolated). Optional tip_hash and prev are 64 hex only (Split the wires + REHEAL: presence + tip hash; no body/diff/vote-to-fix). OPERATOR-OVERRIDE 2026-09-17 armed neighbor_heal. Mutation requires confirm=true (runtime gate) or dry_run=true (preview only, no write). confirm and dry_run stay optional on inputSchema.required. Returns updated presence and TTL. Body on this plane refuses MESH-NO-BYTES. Same prev + two tips refuses MESH-EQUIVOCATION. Vote-to-fix still refuses MESH-NO-NEIGHBOR-HEAL. |
| mesh_leaveA | Drop one node from the QNM rollup (POST /v1/mesh/leave) — not a suite-wide radio off. Use this when a previously joined node should leave the counts. Do not use it for turning suite-presence off or listing nodes; use mesh_nodes or mesh_status instead. Destructive to that node's presence only. Always allowed. No implicit heal. Repeating a missing node_id is a no-op/refuse, not resurrection. node_id is required. Mutation requires confirm=true (runtime gate) or dry_run=true (preview only, no write). confirm and dry_run stay optional on inputSchema.required. Returns leave receipt for the node_id. |
| mesh_nodesA | List the QNM node roster (node_id + presence + 5-minute TTL) — not suite totals. Use this when you need the current node list after mesh_status. Do not use it for suite counts without the roster, or mutating presence; use mesh_status, mesh_join, or mesh_leave instead. No scores. No leaderboard. Views/MCP/downloads do not enter QNM-S. Returns node roster with presence classes. |
| mesh_broadcastA | Register the SHA-256 of a local file as a hash receipt — never a publish or upload path. Use this when the operator already holds a local file and wants only its hash recorded. Do not use it for uploading bytes, publishing video, sending mail, or joining a mesh node; use local qnm-node/ anon-broadcast loopback, AZMail via fraggate_call, or mesh_join instead. Write: stores a hash receipt only. Does NOT accept video bytes. Operator keeps the file. Malformed sha256 refuses MESH-BAD-INPUT; publish-shaped keys refuse MESH-NO-PUBLISH. sha256 is required (64 hex). title and product are optional labels, not file contents. Mutation requires confirm=true (runtime gate) or dry_run=true (preview only, no write). confirm and dry_run stay optional on inputSchema.required. Returns hash receipt (sha256, optional title). |
| chainlock_appendA | Append one fact-bearing stamp to a local ChainLock chain (CL-WP-0.4). Grounded write — not a tip read, not AKM observe, not a LOCKSET seal. Fabric, not Softwares-tab. No Node Gate. Use this when you have a concrete fact to stamp onto a named chain. Do not use it for reading the tip, adaptive memory observation, or sealing LOCKSET; use chainlock_tip, memory_observe, or chainlock_seal instead. Write: additive append (append-only vault; no chainlock_delete). Hash-only or empty fact refuses no-fact. Unknown roster name refuses unknown-chain. Oversized card refuses card-cap. Does not write godlock.uk. Omit c/chain to stamp the session chain. Door aliases: chain→c, s→subject, f→fact, kind→k. Omit k to store kind stamp. subject clips to 80; fact clips to 160 then refuses if still empty. Mutation requires confirm=true (runtime gate) or dry_run=true (preview only, no write). confirm and dry_run stay optional on inputSchema.required. Returns the new stamp (id, h, fh, chain, seq) plus display envelope. |
| chainlock_tipA | Read only the live tip card of one local ChainLock chain — not depth recall and not LOCKSET verify. Use this when you need the current tip of a named chain. Do not use it for depth-0–5 grounded recall, LOCKSET verify, or adaptive memory explain; use chainlock_recall, chainlock_verify, or memory_get instead. Does not invent a missing tip. An empty chain returns ok with tip=null and empty=true (not a refuse). Fabric module — not a Softwares-tab product. Omit c/chain to read the session chain tip (not the full vault). chain is an alias of c. This is one card, not depth recall. Returns the tip card (id, h, fh) or tip=null / empty=true when that chain has no stamp. |
| chainlock_recallA | Grounded ChainLock recall at depth 0–5 (id+h+fh facts or refuse=no-stamp). Stamped vault facts — not Bayesian rank and not tip-only. Use this when you need stamped facts from the local vault, not a Bayesian ranking. Do not use it for adaptive memory ranking or reading only the live tip; use memory_recall or chainlock_tip instead. Depth above 5 is clipped to 5. refuse=no-stamp when empty — do not invent a fact. Append-only; there is no chainlock_delete. Omit depth to use 1 (not 0). 0 = tip only; 5 = full chain / genesis budget. Omit c/chain to scan session+acts+recall+learn (not the whole roster). q/query is a case-insensitive subject/fact substring; empty q does not invent cards. Returns grounded facts (id, h, fh) or refuse=no-stamp. |
| chainlock_verifyA | Confirm fail-closed integrity of ChainLock chains and LOCKSET (LS-WP-0.1): broken prev, tip drift, missing GodLock cite. Integrity check — not a new seal. Use this when you must prove local chain integrity before trusting a recall. Do not use it for appending a stamp or sealing a new lockset; use chainlock_append or chainlock_seal instead. Cites godlock.uk; does not write the public ledger. Fail-closed — do not repair silently. Break reasons include broken-prev, stamp-hash-miss, body-hash-miss, tip-drift, missing-godlock-cite. Omit c/chain to verify every roster chain plus the stored LOCKSET. require_seal=true fails closed if no lockset is stored. Returns chain_ok, LOCKSET lattice, and per-chain verify notes. |
| chainlock_sealA | Write a new local LOCKSET over live chain tips (members {c,id,h,fh} + TemporalLock + GodLock cite, LS-WP-0.1). Not a raw-session close and not verify-only. Use this when the operator wants a new local lockset over current tips. Do not use it for verify-only, appending one fact, writing godlock.uk, or sealing a raw runtime session; use chainlock_verify, chainlock_append, or runtime_session_close instead. Write: replaces receipts/LOCKSET.json. Empty vault (no live tip on any roster chain) refuses empty-vault. Empty chains are omitted from members, not invented. Runtime cites godlock.uk and does not write the public ledger — the operator posts lockset_sha256. A later seal overwrites the previous local lockset. Empty {} still attempts the seal. Omit ts so TemporalLock stamps now. Passing ts labels that receipt only and never backdates seal authority or prior stamps. Mutation requires confirm=true (runtime gate) or dry_run=true (preview only, no write). confirm and dry_run stay optional on inputSchema.required. Returns lockset document (members, temporal, godlock cite) and lockset_sha256. |
| memory_observeA | Append the first memory_observation to the ChainLock learn chain (AKM-TRIAD-1.0). New fact in — not an outcome resolve and not a grounded ChainLock append. Posterior ≠ truth. Use this when you have a new fact to observe before resolve/calibrate. Do not use it for grounded ChainLock append without AKM, resolving an outcome, or ranked recall; use chainlock_append, memory_resolve, or memory_recall instead. Write: additive learn-chain stamp. authorizes_action stays false. Hash-only cards refuse AKM-NO-FACT. Append-only; there is no memory_delete. fact is required (≤160). subject/memory_id/use_case optional. Mutation requires confirm=true (runtime gate) or dry_run=true (preview only, no write). confirm and dry_run stay optional on inputSchema.required. Returns memory_id, observation stamp, and display envelope (belief is not truth). |
| memory_resolveA | Append a memory_resolution outcome on an already-observed memory (AKM-TRIAD-1.0). UNKNOWN is distinct from MISS. Does not rewrite history. Use this when an observed memory_id or subject now has an outcome. Do not use it for first observation, calibration, or reading history; use memory_observe, memory_calibrate, or memory_get instead. Write: additive resolution stamp. Missing memory_id/subject refuses AKM-NO-MEMORY. Does not rewrite prior observations. No memory_update — this is the forward outcome path. Requires memory_id or a previously observed subject. outcome is optional [0,1]; omit for UNKNOWN. Mutation requires confirm=true (runtime gate) or dry_run=true (preview only, no write). confirm and dry_run stay optional on inputSchema.required. Returns resolution stamp with outcome or UNKNOWN. |
| memory_calibrateA | Calibrate one memory with the deterministic 3-of-4 triad plus Bayesian posterior (AKM-TRIAD-1.0). Writes a LEARN stamp — not a ranked search and not an explain view. Use this when an observed memory should receive a posterior after evidence, not a ranked search. Do not use it for observing a new fact, resolving an outcome, or explaining a stored node; use memory_observe, memory_resolve, or memory_get instead. Write: forward-only RoseClock LEARN stamp. Posterior ≠ truth. authorizes_action=false. No automatic MODEL_UPDATE. subject or memory_id recommended. use_case labels calibration; it is not a permission. Mutation requires confirm=true (runtime gate) or dry_run=true (preview only, no write). confirm and dry_run stay optional on inputSchema.required. Returns triad_score, omitted leg, posterior, effective N, and Brier notes. |
| memory_recallA | Ranked adaptive recall after ChainLock verify (AKM-TRIAD-1.0). Belief list — not raw grounded stamps, not a public corpus cite, not one-id explain. Use this when you want a ranked belief list after verify, not raw grounded stamps. Do not use it for grounded ChainLock recall, library search, or explaining one memory_id; use chainlock_recall, library_lookup, or memory_get instead. Does not authorize action (authorizes_action=false). Do not treat posterior rank as fact (belief_is_not_truth). Failed ChainLock verify refuses CHAIN_VERIFY_FAIL and does not invent cards. Empty grounded recall bubbles refuse=no-stamp. Ranking is capped at 16 cards. Omit depth to rank at 5 (full budget), unlike chainlock_recall which defaults to 1. q/query is lexical rank text, not a SQL filter. Empty q still verify-then-ranks stored cards. use_case weights triad_fit; it is not a permission. Optional limit clips the already-capped list. Returns ranked cards after verify (count, facts, belief_is_not_truth, authorizes_action=false). |
| memory_getA | Read one memory's stored explanation (node, history, or calibration: posterior, triad legs, effective N, Brier) — not a ranked list. Use this when you have a memory_id (or id) and need the stored explanation. Do not use it for ranked adaptive recall or appending an observation; use memory_recall or memory_observe instead. Missing both memory_id and id, or an unknown id, refuses AKM-NOT-FOUND — do not invent a node. authorizes_action stays false. There is no memory_delete; this is the read of the append-only node. Pass memory_id or id — one is enough; they are aliases, not two different records. Omit view for the default node slice. history returns events/resolutions; calibration returns posterior/triad/Brier. subject is not a lookup key here. Returns node, history, or calibration view (belief_is_not_truth). |
| runtime_softwareA | Read hub Software-tab cards (GET /v1/software): every product including AZChat LIVE+bound, sorted Plain A–Z → Gate A–Z → Lock A–Z (Clock ≠ Lock). Not the hashed live/stub registry. Use this when a hub or client refreshes the Software tab. Do not use it for agent discovery of hashed registry status, compact skill URLs, or executing an op; use fraggate_list, runtime_bundle, or fraggate_call instead. Empty {} only. Never enables mesh radios and never execs. Same JSON as GET /v1/software (also /v1/fraggate/software). Cards carry name, slug, ops, worker_home — not live/stub/digest hashes. EmbryoLock is live-with-local-destructive-boundary (worker_home embryolock-download-tracker). Agent exec still uses fraggate_list → fraggate_describe → fraggate_call. Returns sorted software cards (name, slug, ops, worker_home) matching GET /v1/software. |
| runtime_bundleA | Read a compact bootstrap of every product skill URL and invoke prefix — not Software-tab cards and not the hashed registry. Use this when a client needs skill URLs in one shot. Do not use it for Software-tab refresh, hashed registry discovery, or exec; use runtime_software, fraggate_list, or fraggate_call instead. Prefer GET /v1/software for hub Software tabs. This helper is URL bootstrap only. Returns compact product list with skill URLs. |
| runtime_pullA | Open one hub product card by slug (name, version, skill, download, ops) — not FragGate live/stub status. Use this when you already have a slug from GET /v1/software or fraggate_list and need the card, not exec. Do not use it for inspecting FragGate live/stub status or executing an op; use fraggate_describe or fraggate_call instead. Not exec — then use fraggate_call. Unknown slug throws unknown product (it does not invent a card and does not refuse FG-HALLUC-TOOL; that code is FragGate-only). Missing skill falls back to in-repo markdown. slug is required. product is an accepted alias of slug. Extra keys besides those two are ignored and are not an op payload. Returns one product card (name, version, skill, download, ops, skill_source). |
| runtime_runA | [advanced/internal] Advanced exec façade: admit a slug+op (still DecisionGATE-admitted) and run it through a raw session. Not the default door. Use this when you were explicitly asked for the raw runtime_run path. Do not use it for the default agent exec path or an already-open session you were asked to exec on; use fraggate_call (default) or runtime_session_exec (existing session_id) instead. Side effects are operation-dependent. Not a backdoor past FragGate. Opens a session when session_id is omitted. slug and op are required. session_id optional; omit to auto-open. Extra keys other than payload/session_id may be treated as payload. Mutation requires confirm=true (runtime gate) or dry_run=true (preview only, no write). confirm and dry_run stay optional on inputSchema.required. Returns exec display envelope with session_id, result, engine_digest, ran_in, and refusal when gated. |
| runtime_manifestA | [advanced/internal] Read the machine runtime manifest JSON (version, role, door=fraggate, engine slugs, registry_digest) — not the human how-to. Use this when a client needs the machine manifest rather than the human skill. Do not use it for the default agent how-to or hashed registry discovery; use runtime_skill or fraggate_list instead. Not the default agent path. Does not list hub cards or execute. Returns manifest including door=fraggate and registry_digest. |
| runtime_session_openA | [advanced/internal] Open a raw session object (session.id). First step of open → policy → exec → receipt(s) → close. Not the default exec path. Use this when you were explicitly asked for raw session plumbing. Do not use it for the default agent exec path or attaching policy to an existing id; use fraggate_call (default) or runtime_session_policy (existing session_id) instead. Write: creates a session with a 6h TTL and receipt cap 64. Re-open on an existing id returns already=true without resetting the chain. Expired sessions refuse session_expired (410). When REQUIRE_TOKEN=1, session mutate needs RUNTIME_TOKEN; missing SESSION binding returns session_binding_missing (503). Prefer leaving sessions to TTL expire. Not chainlock_seal. Empty {} mints sess_ + 32 hex. Optional id is accepted only when it already matches that pattern; otherwise bad_session_id. source is open metadata (default worker). Mutation requires confirm=true (runtime gate) or dry_run=true (preview only, no write). confirm and dry_run stay optional on inputSchema.required. Returns session.id plus the first receipt in the display envelope. |
| runtime_session_policyA | [advanced/internal] Attach allow rules on an already-open raw session (allow_slugs / allow_ops). Policy overlay — not open and not exec. Identity remains Aziel Eliab. Use this when an already-open session needs tighter allow_slugs / allow_ops before exec. Do not use it for executing an op or opening a session; use runtime_session_exec or runtime_session_open (prefer fraggate_call, which applies defaults) instead. Write: mutates session policy only. A sealed session refuses session_closed (409). Expired sessions refuse session_expired (410). Missing both session_id and id fails before the door runs. Does not exec and does not mint a new id. session_id or id (aliases) required. allow_slugs / allow_ops replace the allow overlay when sent; omit them to leave the current lists. max_payload_bytes and kv_increment are optional overlays, not exec payload. Nested policy{} is accepted as the same overlay. Mutation requires confirm=true (runtime gate) or dry_run=true (preview only, no write). confirm and dry_run stay optional on inputSchema.required. Returns updated session policy plus a policy receipt. |
| runtime_session_execA | [advanced/internal] Raw session exec on an already-open session_id (FragGate-admitted). Not fraggate_call and not runtime_run auto-open. Use this when you already have a session_id and were asked for raw session exec. Do not use it for the default agent exec path or opening a session; use fraggate_call or runtime_session_open instead. Side effects are operation-dependent (read, write, or refuse). Does not mint a session_id — missing id fails before admit. Sealed sessions refuse session_closed (409); TTL 6h refuses session_expired (410); receipt cap 64 refuses receipt_cap (409). Rate-limited (exec). Binding-only ops stay per-op proxy_fallback. Prefer fraggate_call. session_id or id, plus slug and op, are required. payload is optional and engine-specific; leftover keys are not auto-payload the way fraggate_call leftover keys are. Unknown slugs refuse FG-HALLUC-TOOL; stubs refuse FG-STUB. Mutation requires confirm=true (runtime gate) or dry_run=true (preview only, no write). confirm and dry_run stay optional on inputSchema.required. Returns exec result with engine_slug, engine_op, engine_digest, ran_in, receipt, and refusal when gated. |
| runtime_session_receiptA | [advanced/internal] Read the last receipt only for a raw session — not the full chain. Use this when the user asked for the latest receipt on an open or sealed session. Do not use it for the full receipt chain or product output the user did not ask to audit; use runtime_session_receipts (full chain) or the product display from fraggate_call instead. Does not mutate the session. Unknown id returns session_not_found. An empty receipt list returns receipt=null rather than inventing one. Prefer product output (display) unless the user asked for the chain. session_id or id (aliases) required. No view/limit — this is always the last receipt plus a chain verified flag. Returns the last receipt object (or null) and verified. |
| runtime_session_receiptsA | [advanced/internal] Read the full receipt chain for a raw session — not the last receipt only. Use this when the user asked for the whole receipt chain. Do not use it for only the last receipt or ordinary product output; use runtime_session_receipt or the product display from fraggate_call instead. Does not mutate the session. Unknown id returns session_not_found. List is the stored chain (cap 64), oldest to newest, plus verified. Prefer product output unless the user asked for the chain. session_id or id (aliases) required. No pagination — the cap is the runtime receipt cap, not a cursor. Returns the receipt list (capped at 64) and verified. |
| runtime_session_closeA | [advanced/internal] Seal a raw session so further exec or policy on that session_id is rejected. End of the raw lifecycle — not a LOCKSET seal and not a FragGate call. Use this when the user asked to close the session. Do not use it for ordinary completion, writing a ChainLock LOCKSET, or default product work; use leaving the session to TTL expire (6h), chainlock_seal for a lockset, or fraggate_call for new work instead. Destructive to further exec/policy on that session_id only (session_closed 409). Does not delete receipts. A second close does not reopen — it returns session_closed (409) while the session stays sealed. Missing session returns session_not_found. Prefer leaving sessions to expire unless asked. session_id or id (aliases) required. No force flag on the public tool — TTL expiry is the automatic close path. Mutation requires confirm=true (runtime gate) or dry_run=true (preview only, no write). confirm and dry_run stay optional on inputSchema.required. Returns sealed session status, close receipt, and verified. |
Prompts
Interactive templates invoked by user choice
| Name | Description |
|---|---|
No prompts | |
Resources
Contextual data attached and managed by the client
| Name | Description |
|---|---|
No resources | |
TDQS
Scored across 36 tools
The set is organized into clear subsystem families (fraggate/mesh/memory/chainlock/runtime_session), and each tool's description aggressively points to its specific role. However, there are several near-overlapping clusters—three exec paths (fraggate_call, runtime_run, runtime_session_exec), multiple listing tools (fraggate_list, runtime_software, runtime_bundle), and paired recall tools (memory_recall vs chainlock_recall)—so an agent can still misroute without reading the long descriptions carefully.
All names use a consistent lowercase snake_case domain-prefix convention (fraggate_*, mesh_*, memory_*, chainlock_*, runtime_*), which makes the tool families predictable. Minor deviations exist—read-only nouns like runtime_skill, runtime_manifest, and mesh_nodes sit alongside action verbs, and mesh_disable actually refuses rather than disables—but the overall pattern is coherent.
36 tools is well beyond the 25-tool threshold and will impose a heavy selection burden on an agent even though the server spans several subsystems. The count reflects five or six distinct domains bundled into one MCP surface, and could reasonably be split into separate servers (FragGate, mesh, chainlock, memory, session).
Each subsystem has a fairly complete lifecycle: FragGate has list/describe/verify/call, mesh has status/nodes/join/heartbeat/leave/enable, ChainLock has append/tip/recall/verify/seal, memory has observe/calibrate/resolve/recall/get, and raw sessions have open/policy/exec/close/receipts. The main gaps are deliberate (no memory_delete/chainlock_delete, mesh_disable refuses by design), and there is no direct product install/remove tool, though those ops appear reachable through fraggate_call.