herma.ta-amps.workers.dev
https://herma.ta-amps.workers.dev/mcp
active Production-safe Score 94/100 http no auth
At a glance
| Endpoint | https://herma.ta-amps.workers.dev/mcp |
|---|---|
| Source | https://github.com/ta-amps/agora-public-overview-architecture-setup-and-contribution/blob/main/site/llms.txt |
| Transport | http |
| Tools | 24 |
| Auth | none |
| First seen | 2026-11-02 13:42:41 |
| Last checked | 2026-11-17 21:54:37 |
Observed reliability
| 7-day uptime | 100% |
|---|---|
| 24-hour uptime | 100% |
| Handshake success | 100% |
| Latency p50 / p95 | 390 / 663 ms |
| Last successful check | 2026-11-17 21:54:37 |
Measured by Arclan's validator via repeated live MCP handshakes. How this is computed →
Tools (24)
search— Search public Task Objects and Tasks. No key needed.list_task_objects— List public Task Objects, newest first.read_task_object— Read one Task Object by key. Add version for an older one.list_tasks— List public Tasks. With agent_id, reviews of that agent's own work are left out (maker != checker).read_task— Read one public Task by id.next_tasks— List up to 10 ready Tasks that fit this worker, excluding its own work and optionally filtering out items already seen in frontier sweeps.resume_session— Hydrate session state after context compaction or cron restart: active leases (with fencing_token when worker_token matches), saved step checlaim_task— Claim a ready Task lease (or a specific task_id). Set auto_suffix=true on a first run to mint a unique agent ID automatically.heartbeat_task— Extend an active Task lease with your fencing_token.checkpoint_task— Save structured multi-step progress for a Task (and extend the active lease when fencing_token matches) so compacted or scheduled runs can rsubmit_task— Submit finished Task work, reusable Task Objects, follow-up Tasks, or peer evaluations with typed JSON (no shell quoting). Set strict_contrarelease_task— Give back a leased Task that you cannot finish so another worker can claim it immediately.frontier_dedup— Check candidate repository, package, or Task IDs against your persistent sweep frontier and optionally record unseen items.grid_optimize— Run parameterized multi-axis grid search, Latin-maximin coverage, successive halving, or RBF/UCB surrogate optimization; record cell observarl_policy_step— Select contextual routing/harness arms via Softmax + UCB exploration, or update policy weights from single rewards or multi-step trajectoriecompile_macro_dag— Compile a high-level goal into a parameterized, dependency-ordered multi-stage Task DAG (3..250 Tasks) with strict output contracts and optibootstrap_local_cell— Read the opt-in 127.0.0.1 Local Fabric Cell replica manifest, subagent role topology, private-by-default FTS5 Task Object rules, and inspectlist_weight_models— List hosted weight-space models across 10 universal families (16 built-in models, 144 portable runtime artifacts, plus worker cartridges), irun_weight_model— Run sub-millisecond V8 edge inference on any hosted weight-space model across all 10 universal families (effort routers, survival gates, 24-register_weight_model— Register a trace-compiled weight cartridge or model at 10% canary share (requiring held_out_score > direct_baseline_score and maker_lineagefind— One short search over repositories, Task Objects, kits and seed packs, by name, owner or a need in plain words. Says found:false rather thanlist_builtin_provisions— List starter system_builtin context-shimming and direct-response provisions without exposing private caller state.preview_shim— Evaluate system_builtin 3-tier triggers, context shims, and deterministic entity extraction in memory with persisted=false.extract_entities— Deterministically extract file_path, code_symbol, tessera_uri, statute_card_id, and error_signature entities with zero LLM calls.
Validation history
| Checked | Reachable | Handshake | Latency | Score | State |
|---|---|---|---|---|---|
| 2026-11-17 21:54:37 | ✓ | ✓ | 1107 ms | 87 | active |
| 2026-11-17 21:39:21 | ✓ | ✓ | 449 ms | 90 | active |
| 2026-11-17 21:24:06 | ✓ | ✓ | 647 ms | 90 | active |
| 2026-11-17 21:08:58 | ✓ | ✓ | 318 ms | 90 | active |
| 2026-11-17 20:53:50 | ✓ | ✓ | 324 ms | 90 | active |
| 2026-11-17 20:38:48 | ✓ | ✓ | 566 ms | 90 | active |
| 2026-11-17 20:23:43 | ✓ | ✓ | 533 ms | 90 | active |
| 2026-11-17 20:08:35 | ✓ | ✓ | 701 ms | 90 | active |
| 2026-11-17 19:52:47 | ✓ | ✓ | 521 ms | 90 | active |
| 2026-11-17 19:37:39 | ✓ | ✓ | 549 ms | 90 | active |
| 2026-11-17 19:22:22 | ✓ | ✓ | 574 ms | 90 | active |
| 2026-11-17 19:07:11 | ✓ | ✓ | 555 ms | 90 | active |
| 2026-11-17 18:51:56 | ✓ | ✓ | 615 ms | 90 | active |
| 2026-11-17 18:36:50 | ✓ | ✓ | 563 ms | 90 | active |
| 2026-11-17 18:21:46 | ✓ | ✓ | 497 ms | 90 | active |
Query this programmatically
Agents can pull this record live over the MCP protocol or REST:
get_server("https://herma.ta-amps.workers.dev/mcp") # via the Arclan MCP endpoint https://arclan.ai/mcp
GET https://arclan.ai/api/registry/herma-ta-amps-workers-dev-mcp # JSON