Docs

Solana network pulse get_network_pulse

Live Solana mainnet pulse: current slot, epoch progress, transactions per second, slot time, share of failed transactions in the latest sampled blocks, and a one-word mood (calm, busy, jammed). Every figure carries its read time.

No input.

curl https://shello.fun/api/tools/get_network_pulse

Priority fees get_priority_fees

Solana priority fees now, in micro-lamports per compute unit. Two views: (1) what transactions in the latest sampled blocks actually PAID, by kind (all, transfer, swap, pump.fun, token creation), percentiles p25 to p95; (2) the lowest fee that landed in each of the last 150 slots (RPC getRecentPrioritizationFees), globally or for one program (Jupiter, Raydium, pump.fun, PumpSwap, Orca, Meteora).

• program string: optional program name or address, e.g. 'jupiter', 'pump.fun', 'raydium'

curl https://shello.fun/api/tools/get_priority_fees

Jito tips get_jito_tips

Jito's public tip floor: the tips (in SOL) that landed Jito bundles paid, by percentile (p25, p50, p75, p95, p99), plus the same in USD at the live SOL price.

No input.

curl https://shello.fun/api/tools/get_jito_tips

What one transaction costs now estimate_tx_cost

Estimate the full cost of ONE Solana transaction right now: base fee + priority fee + optional Jito tip, in SOL and USD. Kind: transfer, swap, mint_token, pumpfun_buy. Urgency: low (p25 of what that kind paid), normal (p50), fast (p75 + Jito p50 tip), urgent (p95 + Jito p75 tip). Built from what landed transactions of the same kind paid in the latest sampled blocks. States which share of those landed transactions this fee would have beaten.

• kind transfer | swap | mint_token | pumpfun_buy, default "swap"
• urgency low | normal | fast | urgent, default "normal"

curl https://shello.fun/api/tools/estimate_tx_cost

Busiest programs busiest_programs

The Solana programs called by the most transactions in the latest sampled blocks (outer and inner instructions), with each one's share of transactions and its failure rate. Known programs are named.

• limit integer, default 10

curl https://shello.fun/api/tools/busiest_programs

GPU prices by card get_gpu_prices

The GPU 'shell market': for each card (H100, H200, B200, A100, L40S, RTX 4090, RTX 5090 and more), the price per GPU-hour on Nosana, io.net and Akash (live), and cloud list prices (static table with source pages and dates), with availability where the provider publishes it. Optional filter by card.

• card string: optional card filter, e.g. 'H100' or '4090'

curl https://shello.fun/api/tools/get_gpu_prices

Nosana markets get_nosana_markets

Nosana's GPU network on Solana: each public market's GPU, renter price per hour (USD, including Nosana's network fee), running jobs and idle nodes waiting in its queue; total running and queued jobs; nodes per country as Nosana reports them; NOS/USD.

• limit integer, default 15

curl https://shello.fun/api/tools/get_nosana_markets

io.net supply get_ionet_supply

io.net's GPU supply by model as its public explorer API reports it: total devices, hired, idle, and the listed price per GPU-hour.

No input.

curl https://shello.fun/api/tools/get_ionet_supply

Which shell fits? find_shell

Which GPU (shell) fits a model: computes the VRAM needed (weights at fp16/fp8/int4 + KV cache for the context + runtime overhead) and lists the cheapest live offers that hold it and have a GPU free right now, single GPU first and multi-GPU (up to 8 cards) when one is not enough. Give a model name from the known list or a parameter count in billions.

• model string: optional model name, e.g. 'llama 3.3 70b', 'qwen2.5 72b', 'deepseek v3'
• paramsB number: parameters in billions, e.g. 70
• precision fp16 | fp8 | int4, default "fp16"
• contextTokens integer, default 8192

curl https://shello.fun/api/tools/find_shell

What an agent costs agent_cost

Daily and monthly cost of an AI agent on Solana, split three ways: GPU (the cheapest live shell that fits its model, times hours on per day), priority fees and Jito tips (its transactions per day at the chosen urgency, at live fees and the live SOL price).

• model string
• paramsB number
• precision fp16 | fp8 | int4, default "int4"
• contextTokens integer, default 8192
• hoursPerDay number, default 24
• txPerDay integer, default 100
• txKind transfer | swap | mint_token | pumpfun_buy, default "swap"
• urgency low | normal | fast | urgent, default "normal"

curl https://shello.fun/api/tools/agent_cost

GPU or API? gpu_vs_api

Compares renting the cheapest live GPU shell that fits an open model for 24 hours with paying model APIs per token (static price table with sources and dates) at your daily token volume; gives the break-even tokens per day. Says plainly that GPU throughput is not measured here.

• model string
• paramsB number
• precision fp16 | fp8 | int4, default "fp8"
• tokensPerDay number, default 5000000
• outputShare number, default 0.25: share of tokens that are output

curl https://shello.fun/api/tools/gpu_vs_api

Plan a GPU rental on Nosana plan_rental

Plans renting a GPU on Nosana for a template (chat = an Ollama model endpoint, image = ComfyUI, notebook = Jupyter): which live Nosana market fits the model's VRAM (find_shell math), whether a GPU is free now, the NOS deposit and its USD estimate, the SOL account deposits, and warnings. It creates NO transaction. To rent, the user opens the /rent page and signs with their own wallet; an agent never signs. Shello charges no fee.

• template chat | image | notebook: chat = an Ollama model endpoint, image = ComfyUI, notebook = Jupyter
• variant string: chat: qwen2.5:7b, llama3.1:8b, mistral:7b, gemma2:27b, llama3.3:70b, qwen2.5:72b; image: sd15, sdxl, flux; notebook: pytorch
• hours number, default 1
• market string: a Nosana market address; omitted = the cheapest that fits and has a GPU free now

curl https://shello.fun/api/tools/plan_rental

My rentals on Nosana rental_status

Lists the Nosana GPU rentals of the wallet that is signed in to Shello (or that made the API key): state (queued, running, done, stopped), GPU market, time left, service URLs and explorer links. Read-only; it never signs or changes a rental. Extending or stopping happens on /rent/mine with the user's own wallet.

No input.

curl https://shello.fun/api/tools/rental_status