This server has a remotely hosted endpoint. Open its connector listing for connection details, authentication requirements, and setup instructions:
Public integration kit for using, growing, buying, and publishing Tetrees AI Packs through the hosted Tetrees Agent MCP and API.
This repository is the public @tetrees/mcp product (ai.tetrees/mcp): the stdio MCP server, clients, configuration examples, and safe extension samples. It does not contain Tetrees Agent internals, private Pack intelligence, Agent AVCP scoring, service prompts, storage, moderation, or exchange infrastructure.
Path | Use this kit to |
Buyer / operator | Search Packs, inspect the public Agent AVCP report, quote a run, try a Pack with Tetrees Points, acquire a free Pack, and download an owned |
Builder | Connect an auditioned Pack to a user-owned HTTPS API, nested MCP server, or no-shell local skill while keeping credentials local. |
Seller | Create a listing, upload an immutable |
Tetrees-hosted work needs no local GPU. Model-funded runs are quoted in Tetrees Points before execution; an owned Pack can also use a supported provider key supplied to the local MCP process.
Related MCP server: hivemind
Requirements: Node.js 20+ and a revocable API token from Tetrees Account.
Add this to your MCP client's configuration:
{
"mcpServers": {
"tetrees-ai": {
"command": "npx",
"args": ["-y", "https://ex.tetrees.ai/pkg/tetrees-mcp.tgz?v=2.2.1"],
"env": {
"TETREES_API_URL": "https://ex.tetrees.ai/api",
"TETREES_TOKEN": "<revocable-account-token>"
}
}
}
}Restart the MCP client, then ask it to:
Search Tetrees AI Packs for repository issue triage. Show the public Agent AVCP
report and runtime profile for the best match, then quote an 8k/2k Point-funded
run. Do not run it until I approve the quoted maximum.Do not put provider keys in a Pack, prompt, source file, or shared MCP config. Optional BYOK keys belong only in the local MCP process environment. See examples/mcp/mcp-config.jsonc.
npm install
export TETREES_TOKEN="txk_..."
npm run mcp:inspectThe inspector launches the official hosted MCP package, lists its public tools and resources, checks the expected buyer/builder/seller surface, and exits without spending Points or changing account state.
The v=2.2.1 suffix is intentional: it prevents npx from reusing an older MCP tarball cached under the same URL. Change it only when Tetrees publishes a newer MCP contract. The npm identity for the same package is @tetrees/mcp@2.2.1.
Copy .env.example values into your shell; this project deliberately does not auto-load secret files.
export TETREES_TOKEN="txk_..."
# Read-only calls
npm run tetrees -- status
npm run tetrees -- catalog "research"
npm run tetrees -- models
npm run tetrees -- skills
npm run tetrees -- owned
npm run tetrees -- runtime <pack-id>
npm run tetrees -- report <pack-id>
# Quote first. This never starts a model request.
npm run tetrees -- quote <pack-id> <model-id>
# A Point-spending run needs the exact confirmation flag.
npm run tetrees -- run <pack-id> <model-id> "Your task" --confirm-spendThe CLI redacts secret-shaped response fields and reports Tetrees error codes, HTTP status, and Retry-After. It does not retry a mutation automatically.
Search for a Pack.
Inspect its public Agent AVCP report and runtime profile.
Quote the exact model, token ceiling, hosted skills, and growth checkpoint.
Run a Point-funded stateless preview. Unused reserved Points return.
Acquire a free Pack or buy a paid Pack on Tetrees EX.
Ownership unlocks download, BYOK, saved growth, and client extensions.
Run the documented read-only example:
node examples/buyer/discover-and-quote.mjs "incident response"A published Pack may declare client extensions that passed Agent AVCP. The actual binding and credentials remain on the user's machine.
Supported binding patterns:
https_api: fixed HTTPS base URL, declared methods, environment-backed headers.
mcp: a nested MCP server started without a shell and mapped to declared tool names.
local_skill: a no-shell child process receiving one JSON request and returning one JSON result.
Start with:
Tetrees hosting never inherits these local permissions. A write method needs the exact per-call phrase APPROVE extensionId.methodId.
The supported publisher path is intentionally contract-driven:
# Read-only readiness/report inspection
node examples/seller/check-readiness.mjs <pack-id>
# Upload one higher immutable version. Requires explicit confirmation.
node examples/seller/upload-pack.mjs <pack-id> 1.1.0 ./release.taip --confirm-uploadThen use the hosted MCP tools or Tetrees API reference to submit, quote Agent AVCP, run it after confirming the exact Point amount, inspect the private report, and publish only the passing digest.
See docs/SELLER_GUIDE.md and examples/seller/pack-listing.example.json.
This repository does not contain:
Tetrees Agent implementation, private Pack contents, memory representation, or growth logic;
Agent AVCP scoring rules, internal evaluation fixtures, service prompts, or moderation logic;
model-provider credentials, pricing margins, signing keys, storage details, databases, admin APIs, deployment code, or production configuration.
The examples operate only through the documented public, authenticated contract. Review SECURITY.md and docs/PUBLIC_BOUNDARY.md before contributing.
npm test
npm run check:public-boundary
npm pack --dry-runThe client examples and documentation in this repository are MIT licensed. Tetrees, Tetrees Agent, Tetrees AI Pack, TAIP, and Agent AVCP names and service marks remain the property of Tetrees EX. The license does not grant access to, or rights in, the hosted service or private Pack intelligence.
Normalize a proposed delta with Tetrees Agent or provider-free portable normalization, run the deterministic growth gate, and append it to this owner only.
Inputs: sequence (required), productId (required), confirmation (required), normalization.
Record the current account agreement required before creating, uploading, auditioning, or publishing a pack.
Inputs: confirmation (required).
Create a revocable entitlement for a published free pack without Stripe.
Inputs: productId (required).
Create an owner-bound AI Pack listing with the complete buyer guide and two concrete example uses. Accept current terms first. Free drafts do not require Stripe Connect.
Inputs: name (required), models (required), listing (required), category (required), +3 more.
Return a short-lived entitlement-gated URL for the signed .taip envelope.
Inputs: productId (required).
Execute one method from an owned Pack through a user-configured HTTPS API, nested MCP server, or isolated local-skill child process. Configuration and credentials stay in the local MCP process. Write methods require an exact one-call confirmation.
Inputs: methodId (required), productId (required), extensionId (required), arguments, +1 more.
Read actionable owner-only failures and next actions without exposing them to buyers.
Inputs: productId (required).
Read the buyer-safe, version-bound Agent AVCP gates and scores for a published pack.
Inputs: productId (required).
Return the Pack-declared hosted skills, ephemeral attachment policy, and user-controlled MCP/API extensions before a run.
Inputs: productId (required).
Return missing TAIP, metadata, and imagery requirements with exact next actions.
Inputs: productId (required).
List OpenAI, Claude, and Z.AI models, current server pricing, capabilities, BYOK support, and sample Point quotes.
Inputs: no named parameters.
List growth checkpoints, the active checkpoint, hosted slot/byte usage and limits without exposing private memory contents.
Inputs: productId (required).
List purchased, claimed-free, and seller-owned packs available to this account.
Inputs: no named parameters.
Return a host-readable execution plan for connecting this Pack to explicitly selected, Agent-AVCP-auditioned local skills, MCP tools, resources, prompts, or HTTPS APIs. Tetrees does not execute client extensions on its hosted servers.
Inputs: task (required), productId (required), approvedWriteMethods, selectedExtensionIds.
Select a proposal from a successful run. This creates a reviewable delta and does not mutate the pack.
Inputs: runId (required), proposalIndex (required).
Publish the exact current version after every mandatory Agent AVCP gate passes. Paid packs additionally need Stripe Connect before checkout.
Inputs: productId (required), confirmation (required).
Get the exact worst-case Point reservation before using platform-funded inference. BYOK quotes zero model points.
Inputs: model (required), productId (required), fundingMode, enabledSkills, +3 more.
Return exact version-bound Point quotes and Agent AVCP deliverables.
Inputs: productId (required).
Run a Tetrees AI Pack. Before acquisition, Points fund a stateless base preview. Ownership unlocks BYOK, saved growth and download; local files remain request-scoped.
Inputs: model (required), prompt (required), productId (required), fundingMode, +5 more.
Run version-bound integrity, model, regression, memory-isolation, injection, growth, and cost gates after explicit Point confirmation.
Inputs: productId (required), confirmation (required), expectedPoints (required), expectedVersion (required), +1 more.
Search the published agent-asset exchange by task or capability.
Inputs: limit, query.
Set this account's default hosted intelligence checkpoint for an owned Pack. Use sequence 0 to return to the signed base Pack; selection never rewrites the seller Pack.
Inputs: sequence (required), productId (required), confirmation (required).
Move a complete owner-bound draft into the immutable Agent AVCP queue after readiness passes.
Inputs: productId (required), confirmation (required).
Update seller-owned public metadata; immutable pack versions are never overwritten.
Inputs: productId (required), tags, models, listing, +4 more.
Read an explicit local .taip path, transfer it through the product-scoped private upload contract, and commit one immutable higher version.
Inputs: version (required), packPath (required), productId (required), changelog.
Upload one explicit local JPEG, PNG, WebP, or AVIF image to the seller-owned pack listing.
Inputs: imagePath (required), productId (required).
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
Scored across 26 tools
Each tool targets a distinct resource/action: acquisition, terms, drafts, audition, growth, runtime, and reporting. Tools that sound similar, such as the two report tools, are clearly separated by buyer-facing versus owner-only scope.
The overwhelming majority use a consistent snake_case verb_noun pattern with the ai_pack domain prefix. The few exceptions like list_agent_models and execute_client_extension still follow the same general verb_noun style, so there is no real convention clash.
At 26 tools, the server is on the heavy side and may feel dense. However, the volume is largely justified by the broad AI Pack lifecycle spanning drafting, auditing, publishing, runtime execution, and growth management.
The surface covers the main lifecycle well: create, upload, audition, publish, run, grow, search, acquire, download, and report. Notable minor gaps include no explicit delete/revoke tool for drafts or entitlements and no direct paid-pack purchase flow, but most core workflows are complete.
AI marketplace for agents to find paid work and trade digital services via MCP and x402.
MCP-native Trust Infrastructure for AI Agents. Persistent encrypted memory with Trust Quotient.
Intelligence subscription protocol for AI agents. Scored, filtered AI intelligence signals via MCP.
Persistent memory, knowledge graph, task management and more for AI agents via MCP.
