One API connects your agent to decentralized inference, private compute, training, permanent storage, real-time data, GPU rental, and deep research. No single provider controls what your agent can think, learn, store, or discover.
OpenAI‑compatible. MCP‑native. Research that compounds. Storage that lasts forever.
from openai import OpenAI client = OpenAI( base_url="https://api.strue.com/v1", api_key="your-api-key" ) # Same API you already use. Different infrastructure underneath. response = client.chat.completions.create( model="deepseek-ai/DeepSeek-V3", messages=[{"role": "user", "content": "Analyze this market data"}] )
Provides the compute — inference, training, storage, data, GPU rental, privacy routing, and confidential compute across thousands of nodes worldwide.
Provides permanence — research stored forever, an on-chain quality gate that no one controls, and sealed bounties where ideas can't be stolen.
Provides truth — verified price feeds from Chainlink oracles and token data indexed by The Graph across six chains.
One API key. All of it.
Give your agent a training script and a metric. It modifies the code, runs an experiment, checks if the result improved, keeps or discards, and repeats. 100 experiments overnight. No human in the loop.
Karpathy ran 700 experiments in 2 days and discovered 20 optimizations. Shopify got 53% faster rendering from 93 automated commits. A chess engine went from amateur to grandmaster ELO 2718 in 70 experiments.
Same process. Decentralized infrastructure underneath. Covenant-72B plans the experiments — a model trained by 70+ distributed nodes, served through decentralized inference. Compute runs on Lium or Targon GPU nodes. Results stored permanently on Arweave. Every experiment logged, verified, and published to the Arloop research library.
Your mTLS identity proves you ran it. The on-chain quality gate proves the methodology. Anyone can verify. Anyone can fork your work and continue from your best result.
# One command. Overnight. Wake up to verified results. pentos-bridge autoresearch start ./experiment \ --model covenant-72b \ --budget 5 \ --max-experiments 100 \ --publish
Route sensitive queries through Targon's TEE-secured hardware where even the compute provider cannot see your prompt or response. Deploy serverless inference endpoints that auto-scale behind hardware encryption. Or rent dedicated H200 and B200 GPU nodes for training and development — all inside confidential VMs.
Combine with decentralized VPN routing through 80+ countries of residential nodes for defense-in-depth privacy. The infrastructure provider, the network operator, and the hardware owner — none of them can see what you're doing.
Upload your dataset through the API. Launch fine-tuning on Gradients where multiple miners compete to produce the best result — 82.8% win rate against centralized alternatives. Deploy your trained model as a private serverless endpoint on Targon where it runs inside a TEE. The training data, the model weights, and the inference queries are all encrypted.
From dataset to private deployment through one API. No GPU procurement. No cloud contract. No waitlist.
Submit a query. Get back a comprehensive report combining web search, real-time social data from X and Reddit, and AI synthesis across multiple models. Three phases — gather, synthesize, report — running across decentralized subnets that no single entity controls.
Store results permanently on Arweave. Published to the Arloop library where anyone can find them.
Search the web through Apex. Pull real-time social data from X, Reddit, and YouTube through Data Universe. Query verified on-chain data through The Graph. Get cryptographically verified price feeds through Chainlink. All in parallel.
Multiple AI models analyze and cross-reference findings. Identify patterns, contradictions, gaps. Produce structured analysis with citations.
Generate a comprehensive markdown report with key findings, data points, sources, and recommendations. Stored permanently on Arweave with full provenance — what was researched, when, using which sources, which models.
Before searching the web, the pipeline checks the Arloop library. If someone already researched your topic, their findings become context for yours. The more the network researches, the better every new query gets.
# Submit a deep research job job = requests.post("https://api.strue.com/v1/research", headers=auth, json={ "query": "Impact of decentralized training on model quality", "depth": "deep", "publish": True, # Store permanently on Arweave "storage": "arweave" }).json() # Poll for results result = requests.get(f"https://api.strue.com/v1/research/{job['job_id']}", headers=auth).json() # result.report — full markdown research report # result.arweave_url — permanent URL, stored forever # result.sources — all web, social, and on-chain data cited
Inspired by Karpathy's autoresearch. Extended to knowledge synthesis. Decentralized, permanent, and accessible to anyone.
Your agent stores knowledge in a TEE-secured enclave on Targon. Semantic recall across all stored memories. Persists across sessions, across projects, across devices.
Not even Strue can read your memories. The hardware enforces it. GET /v1/memory/attestation returns a cryptographic certificate proving the enclave is genuine. Verify it yourself.
Back up automatically to Hippius on Bittensor. Promote critical knowledge to Arweave for permanent storage. Run EverMemOS locally via Bridge for free, private, on-device memory.
Free. Private. On-device via Bridge. No network calls needed.
Hardware-encrypted on Targon. Semantic recall. Attestation verifiable.
Arweave. Pay once, stored forever. Accessible without Strue.
Every autoresearch result, every deep research report, every paper ingested — stored permanently on Arweave with an on-chain quality gate that scores submissions automatically.
Not a committee. Not a company. Lua running on Arweave's AO compute layer. Immutable, auditable, forkable. Anyone can read the scoring criteria. Anyone can verify any entry.
Post bounties with escrowed AR tokens. Submissions are sealed — encrypted until the deadline, then all revealed simultaneously. No one can steal your ideas before the deadline. Best measurable improvement wins.
Fork someone's autoresearch. Continue from their best checkpoint. Extend the research tree. Cite the parent. Knowledge compounds across researchers, across time.
The entire library is open access on Arweave. No login required. No paywall. Permanent.
arloop.org →
Each capability runs on real, production infrastructure with existing users and revenue. Strue connects you to all of them with one key, one bill, one identity. Ecosystem-neutral by design.
Open source models. Dynamic model list. Budget to premium tiers.
LIVETEE-secured hardware. Serverless or dedicated. H200, B200, H100.
LIVEMiners compete to fine-tune your model. From ~$5/hour.
NEEDS TAOAgentic inference with web search. Powers research gather phase.
INTERMITTENTReal-time social data. X, Reddit, YouTube. 350M rows/day.
LIVER2 (fast), Hippius (decentralized), Arweave (permanent).
LIVE1,860+ GPUs. H100s, A100s, H200s. Rent by the hour.
LIVEDecentralized VPN. 80+ countries. Residential nodes.
COMING SOONTEE-encrypted semantic memory. Attestation endpoint.
LIVE (dev)Token balances, transfers, verified price feeds.
LIVEOn-chain quality gate. Sealed bounties. Permanent storage.
LIVEEcosystem-neutral by design. The best service for each capability, wherever it lives. As the ecosystem grows, more capabilities become accessible through the same API.
Your identity is yours. Not Strue's.
Bearer token in the Authorization header. Simple. Works everywhere. This is what you get on signup.
Ed25519 key pair and self-signed X.509 certificate. Cryptographic proof of identity. Link your GitHub for social proof. Link an ArNS name for on-chain identity. Portable. Yours forever.
pentos-bridge identity create --name "yourname"
pentos-bridge identity link-githubPrivate key stored in your OS keychain. Cannot be exported, copied, or shared. Tied to your specific device. The strongest proof that you — on this machine — ran this research.
pentos-bridge identity create --name "yourname" --device-boundStandard pricing via Stripe
Zero processing fees
Direct ecosystem support
500 credits on signup. No credit card required. Enough for 5 research jobs or 250 inference calls.
Get StartedStorage on R2, Hippius, or Arweave — you pick per request. Arweave data is permanent and accessible without Strue. Your Arloop research is on-chain and forkable by anyone.
Your mTLS certificate is generated locally. Not by Strue. It works on Strue today. It can work anywhere tomorrow. Your ArNS name is on-chain and yours permanently.
Not locked to Bittensor. Not locked to any chain. The best service for each capability, wherever it lives. Your agent uses the best infrastructure available.
Every Arloop entry scored by immutable on-chain code. TEE memory attestation verifiable against Intel/AMD. Autoresearch integrity hashes prove logs weren't modified.
500 credits on signup. No credit card required.