The New Tripoli MCP server exposes the same physics kernels that power this site's sim catalogue — as MCP tools any compliant assistant can call directly. Every run is client-computed and emits an OpenChainGraph-conformant artifact carrying a verifiable SHA-256 execution_hash.
A stateless Cloudflare Worker speaking the MCP 2025-11-25 Streamable-HTTP transport. No sessions persist between calls; nothing about a run is stored server-side beyond the response it returns.
Zero-egress: all tool logic runs the same client-side compute the sim pages use — the MCP server returns deep-links and hash-verifiable metadata, never scenario data pulled from anywhere else. No PII in, no PII out.
Pick whichever your client supports — all four point at the same endpoint.
Settings → Connectors → Add custom connector → paste the URL below.
https://mcp.newtripoli.xyz/mcp
Add to ~/.cursor/mcp.json:
{
"mcpServers": {
"newtripoli": {
"url": "https://mcp.newtripoli.xyz/mcp"
}
}
}
Add to .vscode/mcp.json (HTTP server type):
{
"servers": {
"newtripoli": {
"type": "http",
"url": "https://mcp.newtripoli.xyz/mcp"
}
}
}
Raw JSON-RPC over HTTP — list every tool:
curl -s https://mcp.newtripoli.xyz/mcp \
-H 'Content-Type: application/json' \
-H 'Accept: application/json, text/event-stream' \
-d '{"jsonrpc":"2.0","id":1,"method":"tools/list"}'
The catalog is 13 kernels + 7 chains today — it grows. Cold agents should resolve everything through these, not a baked-in list.
| tool | what it returns |
|---|---|
list_newtripoli_tools | Every registered tool: id, mandate, register, guest-legal flag, one-line description — read live off the tools manifest. |
find_tool | Free-text query → best-matching tools (BM25-ranked over titles/descriptions/citations). |
find_chain | Free-text query → best-matching chains, with step order and gate summary. |
build_workflow_links | An ordered set of deep-links for a named chain, inputs preset per step. |
run_chain | Executes a named chain end-to-end server-side, threading each step's execution_hash into a composite hash; stops at the chain's gate if one trips. |
verify_execution_hash | Recomputes a hash from a supplied {policy_parameters, output_payload} pair and reports match/mismatch — the tamper check. |
Counts drift on purpose: alt-history kernels are queued and will move this from 13→18 tools without changing this page's contract. Always call list_newtripoli_tools / find_tool rather than assuming a fixed set.
| tool | register | guest | description |
|---|---|---|---|
nt_time_dilation | canon | yes | Subjective-years-lived multipliers for a given dilation rate and Universal Sabbath reset interval. |
nt_kinetic_probe | canon | yes | ETI kinetic-probe delivery: travel time, relativistic cruise KE, and the vaporization margin gating the deceleration lottery. |
nt_vat_feasibility | feasibility | no | Total power draw of a biological/digital population split, Landauer-priced, vs the Sahara solar cap. |
nt_acceleration_ceiling | feasibility | yes | Subjective-time acceleration ceiling by augmentation stage — floor, ceiling, midpoint, thermal load, named bottleneck. |
nt_comms_lag | canon | yes | Round-trip comms lag between two parties running at different subjective-time rates. |
nt_ring_density | canon | yes | Orbital ring + core habitat housing capacity given depth, area allowance, floors, core population. |
nt_birthday_sacrifice | canon | yes | Subjective time cost of a dilation-rate gap between you and a family member, in Tripoli-block units. |
nt_synthetic_body | canon | yes | Mass, human-mass ratio, and water-buoyancy of a synthetic body skeleton scaled from the canon osmium reference build. |
nt_selection_cost | canon | yes | How many of the 8.1B canon population are excluded/remain under a named upload-selection criterion. |
nt_interface_bandwidth | real-science | no | How far a given brain-computer interface channel count is from whole-brain (~86B neuron) bandwidth. |
nt_tech_tree_path | canon | yes | Solves the post-Wake reindustrialization dependency graph (35 nodes, tiers T0–T9) to a femtoscale assembler. |
nt_provenance | real-science | yes | Emits a tamper-evident ChainGraph provenance manifest for an arbitrary sim run, optionally threaded to a parent hash. |
nt_feasibility_crosswalk | feasibility | no | Grades the ten load-bearing feasibility claims (C-D1..C-D10) against one scenario config; emits verdict ledger + binding verdict. |
Each chain sequences 2–4 tools into a single argument. Call run_chain with the chain name to execute it end-to-end and get back a composite execution_hash.
The whole scenario as a delivery→power→dilation→thermal pipeline; terminal verdict = weakest link.
The D1 delivery fix as a fast-fail decision.
The D10 → D7 hinge.
"Can it actually run": power → interface → thermal.
The physics of friendship: two dilated parties at different rates, plus the comms budget across the gap.
The OCG backbone every other chain terminates into: any tool's run, threaded to a provenance-manifest anchor.
The meta fan-in: grades all C-D1..C-D10 claims against one scenario config, then anchors provenance.
Every run emits an OpenChainGraph v0.4 artifact carrying a SHA-256 execution_hash over {policy_parameters, output_payload}. Same inputs → same hash, deterministically; change one input and it changes. verify_execution_hash recomputes and re-checks it — anyone can audit a claimed result without trusting the party who ran it.
Chains thread each step's hash into a composite_execution_hash, so a multi-tool argument is verifiable as one unit, not just its parts. See the provenance instrument for a live, minimal version of this in the sim catalogue, and the machine-readable catalog for the full tool+chain graph.
CORS allowlist: newtripoli.xyz, www.newtripoli.xyz, claude.ai, app.claude.ai, plus localhost dev ports. No cookies, no session storage — the transport is stateless per request.