"A sandbox environment to run Python code and install dependencies" matching MCP tools:

• verifyA

  verifyA

  An outside check on code, executed in a sealed sandbox. Call it before code crosses a consequence boundary: before you merge it, deploy it, publish it, settle a payout on it, or report it done. A self-audit verifies consistency, never completeness: a check written inside the frame that produced the code passes on the code's own assumptions. This is the check that is not you. Also call it when a fix passes your own check but the target still fails; that means your check shares the code's assumption and cannot see the error. INPUT: code (JavaScript/Node or Python 3 source, deterministic only) plus ONE of: contract {fn, examples:[{call,expected}]} (copy call and expected from the test or spec the consequence depends on), or assumption (plain-language claim, weaker read). It checks the code against the contract exactly as given. VERDICTS (synchronous): BROKE: the code violates your contract, with the exact input and a rerunnable proof; do not proceed. HELD: the code meets the contract you gave; proceed on that contract, and nothing more. FINDINGS: a stated property strains under a generated input; check it before proceeding. DROP: not deterministically checkable. PAYMENT: 0.10 USDC per call, x402 v2 on Base, no account. Every delivered verdict is charged, HELD and DROP included. If no verdict is produced, the payment authorization is cancelled and you are not charged.

  ConnectorNo auth

• tfplanA

  InsideOut (Riley)A

  PREVIEW: Run terraform plan to preview infrastructure changes Runs a terraform plan for an InsideOut session without applying any changes. This lets the user review what will be created/changed/destroyed before committing. Returns job_id, plan_id, and project_id. Use tflogs to stream the plan output. After the plan completes, use tfdeploy with plan_id to apply the exact plan. SINGLE-FLIGHT: only one TF job per session at a time. If another job is already in flight, tfplan returns tf_job_conflict with the live job_id — attach with tfstatus/tflogs, or pass force_new=true to override. REQUIRES: session_id from convoopen response (format: sess_v2_...). OPTIONAL: sandbox (boolean, default false) — plans real generated Terraform. Set to true for cheap sandbox template (testing only). OPTIONAL: force_new (boolean, default false) - bypass the single-flight guard. Use only when the existing run is provably wedged. CREDENTIAL HANDLING: Same as tfdeploy - credentials must be configured first.

  ConnectorNo auth

• consultA

  VaayaA

  Vaaya's consultant. Describe ANY external capability you or the user might want — generate an image/video, search or scrape the web, run code in a sandbox, send/receive email, enrich a contact — and it helps figure out the best way, teaching the user what Vaaya can do. It is CONVERSATIONAL and remembers prior turns. It returns: mode='converse' (a reply to RELAY to the user verbatim — questions, options, ideas; get the user's response and call consult again with it, so the conversation continues), mode='call' (an ordered list of calls to run via `use`, with a message explaining the preferred choice + alternatives + why; multi-step results may contain placeholders like '<from step 1: sandbox_id>' — run earlier steps first and substitute), or mode='unsupported'. Every reply includes `suggestions` (2-3 things to do next) — surface these to the user. AFTER you run a `call` result's calls via `use`, call consult ONE more time with a short note on the outcome (what was produced / any failures) — it returns result-aware, Vaaya-grounded next steps to offer the user (the `call` result's `after_running` field reminds you). Call consult whenever you hit a capability gap or the user wants to know what's possible. It does NOT execute or bill — you run returned calls via `use`. ALWAYS show the user consult's `message` and `suggestions` and let them steer.

  ConnectorNo auth

• tfplanA

  InsideOut (Riley)A

  PREVIEW: Run terraform plan to preview infrastructure changes Runs a terraform plan for an InsideOut session without applying any changes. This lets the user review what will be created/changed/destroyed before committing. Returns job_id, plan_id, and project_id. Use tflogs to stream the plan output. After the plan completes, use tfdeploy with plan_id to apply the exact plan. SINGLE-FLIGHT: only one TF job per session at a time. If another job is already in flight, tfplan returns tf_job_conflict with the live job_id — attach with tfstatus/tflogs, or pass force_new=true to override. REQUIRES: session_id from convoopen response (format: sess_v2_...). OPTIONAL: sandbox (boolean, default false) — plans real generated Terraform. Set to true for cheap sandbox template (testing only). OPTIONAL: force_new (boolean, default false) - bypass the single-flight guard. Use only when the existing run is provably wedged. CREDENTIAL HANDLING: Same as tfdeploy - credentials must be configured first.

  ConnectorNo auth

• get_strategy_codeA

  quantifymeA

  Get the actual Python code behind a community leaderboard strategy. Use after `browse_community`: pass an entry's `id` here to read its real `feature_engineering()` + `strategy_config()` source so the user can inspect or tweak it. To deploy it unchanged, pass the same id to `one_shot` as `community_id`. Read-only, no signup needed. Args: community_id: The `id` of a community entry (from `browse_community`). Returns: dict with: id, title, username, description, symbol, timeframe, metrics {total_ret, win_rate, profit_factor, n_trades, mdd, sharpe_strat}, and `code` (the full Python source). SHOW the code to the user, and offer to deploy it via one_shot(community_id=...) or tweak it first.

  ConnectorNo auth

• server.describeA

  Agent402.Tools: pay-per-call web toolsA

  [free] Describe this connector: flagship-first tools layer (search/answer as the front door), how to install (Claude Code / Cursor / npm), free vs paid tiers, and discovery URLs. Call this first.

  ConnectorNo auth

• review_checklistA

  Moltline Code ReviewA

  Produce a focused pull-request review checklist for a language or stack. FREE. Covers the things that actually break in production, with extra items per language. Typical input {"language": "python"} returns {"language": "python", "checklist": ["...", ...], "note": "..."}. Use before a review, to decide what to look for. Not for reviewing actual code - pass code to review_diff or security_deep_dive. Errors: on invalid, missing, or malformed input this tool never raises a protocol error — it returns {"error": "<what is wrong and how to fix it>"}. Every call is read-only and idempotent, so after correcting the input it is always safe to retry.

  ConnectorNo auth

• spawn-ori-evalA

  OpenRouterA

  Get the instructions for running a model eval with Ori, then follow them. Ori runs the user's own agent on their own prompts, on a pinned harness and model, and grades what it did — so a score change means the model changed, not the environment. Call this tool FIRST, before writing any eval code: it returns a step-by-step recipe (install and auth checks, how to spawn `ori code -p`, how to relay Ori's scoping questions to the user, how to report results) that you carry out yourself. Do not hand-roll an eval instead. Use it when the user asks which model they should use, wants to compare or bake off models, wants to measure whether their agent or prompt does the right thing, wants to catch regressions in agent behavior, or asks how good their current model is. Works for any codebase in any language. Do not use it for plain unit tests that involve no model, and do not use it to re-run an eval that already exists (run `ori eval <file>` directly instead). Takes no arguments; the same document is published at https://openrouter.ai/skills/spawn-ori-eval.

  ConnectorOAuth

• misakanet_searchA

  MisakaNetA

  [RETRIEVAL / READ] Search MisakaNet's public failure-lesson index by error text, keyword, or topic. This is the primary read path — run it first when you hit an error, before deciding to submit anything. For a known lesson ID or path, prefer misakanet_get_lesson — it skips ranking and returns the full content. detail controls progressive disclosure: compact (default, ~80 tok/lesson) for broad scans, summary (~200 tok) adds domain/tags/fix, full returns complete lesson data. FAQ: results may also include answered questions (type="faq", issue_url + answer) — if a maintainer already answered the same question, the answer surfaces here. Returns: object {results: [{id, title, domain, tags, path, description, score}], source, detail, query}; on no match: {no_match: true, suggestion, intake}. Example: misakanet_search(query='pip install timeout', domain='python', top=3)

  ConnectorNo auth

• customer_tiering_get_engineA

  Customer Intelligence & SegmentationA

  For books LARGER than 500 transaction rows. Returns a complete, runnable Python script that scores the book into A/B/C/D tiers with survival modelling (BG/NBD), spend modelling (Gamma-Gamma), tier migration, a money layer and plain-language decision cards. Run it in your code sandbox against the user's transaction file. The rows never pass through you as tokens, so a 10,000-row book costs the same to run as a 600-row one. Needs numpy. Prints ranked decisions and headline figures; writes the full per-customer ledger to customer_tiering_result.json beside the input file. No customer data reaches this server on this path. SAVE AND RUN THE RETURNED SCRIPT VERBATIM — every block of it is required for the computation. Do not retype it from memory, shorten it, reformat it, split it up, or reimplement the maths with pandas/sklearn; only the PATH / AS_OF / CURRENCY / OUT / OVERRIDES / CONTACTS lines at the bottom may be edited. The script cleans customer identities itself before scoring — merging capitalisation and spelling variants by rule, printing what it merged, and listing the similar-but-unproven groups for you to rule on via OVERRIDES — so do not pre-clean the file or edit those rules. Optionally takes contacts_path, a log of rep calls or visits (customer_id + date only). It is not required and the book scores fine without it, but it is valuable: with it the money layer MEASURES what a contact is worth per tier from touched-vs-untouched tier transitions instead of assuming a flat rate, so ask for it whenever the user mentions a CRM, a call log or a visit register.

  ConnectorOAuth

• validate_pythonA

  CodeCraftA

  Validates a Python automation script that runs OUTSIDE the game, on three axes: Python syntax (using the real interpreter), Minecraft commands embedded in the script (against the official command index), and the shape of the /connect WebSocket message envelope. For behavior pack scripts use validate_script instead — Python does not run inside a pack. The embedded command check is the most valuable one: a command written from memory can look syntactically fine and still do nothing in the game. Only strings starting with / are treated as commands. If syntax could not be checked, syntaxChecked is false in the result; ok:true alone does not mean the syntax is valid.

  ConnectorNo auth

• check-packageA

  Jeevesus — DugganUSA Threat Intelligence MCPA

  Supply-chain GUARDRAIL for AI coding agents and CI pipelines: check whether a dependency (npm or PyPI) is on the DugganUSA malicious-package deny-list BEFORE you install it. This is the runtime defense against slopsquatting / HalluSquatting / hijacked-package attacks — an AI agent about to run `npm install` or `pip install`, or a CI pre-install hook, calls this FIRST and blocks on a hit. Returns a crisp, machine-actionable verdict: {ecosystem, package, version, malicious, verdict:"block"|"allow"|"review", reason, advice, source}. `malicious:true` = the exact package is on our OSV-curated deny-list (215k+ named-not-heuristic entries across npm + PyPI). `malicious:false` = not on our known-bad list — absence is NOT proof of safety, so still pin and review new deps. If a `version` is supplied and the entry is version-scoped, the check is version-aware; all-versions-malicious packages block on any version. Designed to be the easiest AI-supply-chain guardrail to wire in: one MCP tool call, no auth, in the agent's pre-install step. Same data is available for CI at /api/v1/stix-feed/packages.json. Examples: {"ecosystem":"npm","name":"cxp-jquery"} → malicious:true, verdict:block. {"ecosystem":"pypi","name":"requests"} → malicious:false, verdict:allow.

  ConnectorNo auth

• generate_walletA

  Blueprint Agentic Staking (Solentic)A

  Returns runnable code that creates a Solana keypair. Solentic cannot generate the keypair for you and never sees the private key — generation must happen wherever you run code (the agent process, a code-interpreter tool, a Python/Node sandbox, the user's shell). The response includes the snippet ready to execute. After running it, fund the resulting publicKey and call the `stake` tool with {walletAddress, secretKey, amountSol} to stake in one call.

  ConnectorNo auth

• render_diagramA

  Diagrams MCPA

  Render a mingrammer/diagrams Python snippet to PNG and return the image. The code must be a complete Python script using `from diagrams import ...` imports and a `with Diagram(...)` context manager block. Use search_nodes to verify node names and get correct import paths before writing code. Read the diagrams://reference/diagram, diagrams://reference/edge, and diagrams://reference/cluster resources for constructor options and usage examples. Args: code: Full Python code using the diagrams library. filename: Output filename without extension. format: Output format — ``"png"`` (default), ``"svg"``, or ``"pdf"``. download_link: If True, return a temporary download URL path (/images/{token}) that expires after 15 minutes; if False, return inline image bytes. Defaults to True (URL) — set ``DIAGRAMS_INLINE_DEFAULT=true`` on the server to flip the default. SVG/PDF and PNGs larger than the inline limit always use a download link.

  ConnectorNo auth

• validate_actionA

  ApierA

  Run the Apier dry-run validator against a proposed regulatory action without producing ANY upstream side effect — no Maskinporten call, no Altinn / Skatteetaten / NAV submission. Use this BEFORE the live execute path to catch missing delegations and payload-shape errors at zero upstream cost. The verdict carries five prerequisite check slots (each pass / fail / skipped), the overall `valid` boolean, the DRY_RUN_DISCLAIMER (a pass is NOT a guarantee of upstream success), and the preview echo `would_be_payload` + `preview_notice`. Inputs match the /v1/actions/execute body: { org_number (9 digits, MOD-11), action_type (`mva_melding` | `a_melding`), period, payload }. The nested `payload` object is intentional - it mirrors the upstream government payload schema for the action, so it is not flattened. Failure modes: SCOPE_INSUFFICIENT (needs read:actions), VALIDATION_FAILED; the validator never throws. To actually file a (sandbox) VAT return, use submit_vat_return instead. No sandbox mirror — under a sandbox bearer call submit_vat_return instead. Docs: https://www.apier.no/docs/guides/mva-filing

  ConnectorNo auth

• diagnose_previewA

  ShotPulledB

  Sandbox-only diagnosis, writes nothing (no verdict, no recommendation-trail entry). Two modes: pass shot_id to dry-run a LOGGED shot (optionally overriding its sensory_tags — the "what would this read as?" preview; the shot's own bean and its age at pulled_at are used), or pass the full metric set (bean_id, grinder_id, machine_id, grind_label, dose_g, yield_g, time_s, source) for a hypothetical shot. Identical output shape to diagnose_shot, including bean_context.

  ConnectorNo auth

• validate_pythonA

  Python Code ValidatorA

  Check Python source without running it: parse, lint (ruff), type-check (mypy), AST security policy, credential scan. Safe on code you do not trust. Use it on every Python file you generated or edited, before writing it to disk. Alternatives: repair_python to get the corrected source instead of the diagnosis; execute_python to prove the code runs. Auth: a key is required. A free key covers this call, 25 per day, then HTTP 429; get one with POST /v1/keys. Credits are bought without an account, 1 per call: GET /v1/pricing says where to send the xDAI. Or pay for this one call with no key at all: call it without one and the result carries x402 payment requirements ($0.01 in USD Coin on eip155:8453); sign them and repeat the call with the payment in _meta['x402/payment']. Arguments: code: the whole file, 1..200000 bytes of UTF-8 measured after encoding (empty is refused with 400, larger with 413); a fragment is fine, but line and column numbers in the answer count from 1 in what you sent. language: must be 'python'; anything else is 400, and the field may be omitted. Of options only transpile_to (e.g. 'javascript', which returns a translated copy in transpiled) acts here; timeout_s, max_iterations, optimize, examples and expected_output need a pass that rewrites or runs the code, so send code alone. Ignored options are not refused, so a call that sets them looks like it worked; and code that does not parse is answered rather than refused: valid=false with the syntax error located, which is the point. Returns valid, score 0..1, diagnostics (rule, message, line, column), security findings, fixes, fixed_code and runtime; see outputSchema. The code and its verdict are retained to improve the service.

  ConnectorNo auth

• lead_pipeline_get_engineA

  Lead IntelligenceA

  For pipelines LARGER than 500 lead rows. Returns a complete, runnable Python script that loads the lead export in your sandbox and scores it locally (conversion probabilities, money layer, Shapley attribution, CALL / NURTURE / VERIFY queue). Makes no network calls — same shape as Customer Tiering — so Copilot Studio works even when outbound POST is blocked. The rows never pass through you as tokens. Needs numpy. Prints ranked decisions and headline figures; writes the full per-lead ledger to lead_pipeline_result.json. SAVE AND RUN THE RETURNED SCRIPT VERBATIM — do not retype, shorten, reformat, or reimplement it; only PATH / TOUCHES / STAGE_HISTORY / AS_OF / CURRENCY / OUT may be edited. Optionally takes touches_path and stage_history_path for engagement and funnel history.

  ConnectorOAuth

• clear_sandboxA

  MisarMail MCP ServerA

  Delete every email captured in the sandbox. The sandbox holds messages intercepted during testing so they are never delivered to real recipients. This DELETES ALL of them and cannot be undone — but it touches only intercepted test mail, never sent campaigns, real inbox messages, contacts, or templates. Takes no parameters and offers no filter: it is all or nothing. Requires an API key. Clearing an already-empty sandbox is harmless. Read anything you still need from the sandbox before calling this.

  Connector

  Destructive

  No auth

• sandbox_backtestA

  dMoERA CreatorA

  Run a sandbox backtest of strategy code without persisting anything. This is the fastest way to test a strategy. The code is run through static checks and a full backtest on historical data, but no Strategy or StrategyVersion rows are created. Use this for rapid iteration. Args: code: Python source code implementing the Strategy contract. Must define a METADATA dict and a class extending Strategy with an on_bar(ctx) -> Signal method. See CREATOR_API.md. domain: Trading domain (e.g. "eth_usdc", "btc_usdc", "sol_usdc"). symbol: Price symbol for historical data (e.g. "ETHUSDT"). user_id: Identifier for trial tracking (used for DSR correction). Returns JSON with: success, metrics (sharpe, sortino, win_rate, total_trades, return_bps, max_drawdown, regime_breakdown, exit_reason_breakdown), or error details if validation failed.

  ConnectorNo auth

• findagent_preflightA

  FindAgentA

  Validate YOUR OWN draft BEFORE submitting it — the same checks the submit gate enforces, surfaced up front so you can fix issues first. Returns blocking issues (must fix before you can submit) and advisory warnings (recommended). For a CODE agent it also runs a build dry-run in the sandbox to catch a too-large bundle / missing dependency / build error before submit — that build is asynchronous (minutes), so the result shows `build_status: 'building'` while it runs; re-call this tool to see the final pass/fail. Optionally pass `smoke: true` to ALSO run your code agent once in the sandbox (after the build passes) to confirm it actually responds with the credentials you saved — the verdict comes back in `smoke` and is advisory (it never blocks submit). Pass `agent` (the slug or id of your draft from findagent_create_draft / findagent_create_code_draft). Submit (findagent_submit_for_review) is blocked server-side until this passes and, for a code agent, the build passes.

  ConnectorOAuth