Grixo dispatches crews of AI agents and puts each execution under instrumentation: step status, streamed output, exact token debit. If it ran, the board shows it and the ledger has the row.
Every dispatch reports step status and streams its output while it runs; every debit posts to an append-only ledger with the balance after it. If it ran, the board shows it — and the ledger has the row.
Stand up the crew — from scratch or from 330+ stored configurations — and assign a model to each operator.
Wire the execution graph and connect the services it reads from and reports to.
Launch from the console, on schedule, or via the REST API. The board tracks every step in real time.
Read the ledger row: exact token debit, balance after. Top up the wallet when reserves run low.
Six panels on the rail. Together they cover dispatch, observation and settlement.
Define operators — agents with roles, goals and models — and slot them into an execution graph on a visual board.
Each dispatched run reports step-by-step: queued, running, done. Output streams to the console as it is produced.
Append-only, integer-exact. Every debit writes the balance after it, so drift is impossible and audits are a read, not a project.
Fire any crew with an authenticated POST. Bearer keys are stored hashed, revocable on the spot, and billed to the same wallet.
120+ connectors put Slack, Telegram, RSS, webhooks and raw HTTP within reach of every operator on the board.
Runs reserve tokens before they start; a failure releases the reserve automatically. Incidents are free — by design, not goodwill.
Power tokens are the only currency on the board. Pick a pack size below — the bigger the load, the lower the per-token rate. Nothing renews.
All packs are one-time purchases. The ledger keeps every movement on record.
Provision an operator and a starter token grant lands on your wallet on entry — enough to dispatch real crews and watch the ledger fill.
Open the consoleRuns climb the board as crews dispatch; each bar is a settled execution with its token debit on the ledger. Impact windows, throughput, spend — all read off one instrumented surface, in real time, never estimated after the fact.
Define operators — agents with roles, goals and models — and slot them into an execution graph on a visual board.
Each dispatched run reports step-by-step: queued, running, done. Output streams to the console as it is produced.
Append-only, integer-exact. Every debit writes the balance after it, so drift is impossible and audits are a read, not a project.
Fire any crew with an authenticated POST. Bearer keys are stored hashed, revocable on the spot, and billed to the same wallet.
Yes. You load a wallet with power tokens; dispatches debit it. There are no seats, no tiers and no renewals — if you stop running crews, spending stops with them.
Integer-precise. Balances and debits are whole tokens, every mutation appends a row carrying the balance after it, and rows are never edited or deleted.
Zero. The engine reserves tokens at dispatch and releases the reserve on failure. Both the reserve and the release appear in the ledger, so the audit trail stays complete.
Yes — register Anthropic, OpenAI, Gemini, xAI or any OpenAI-compatible keys and assign models per operator. Keys are AES-encrypted at rest.
Every crew has a REST endpoint. Authenticate with a bearer API key — stored hashed, displayed once, revocable instantly — and the run bills to the same wallet as console dispatches.
Service credentials are AES-256 encrypted and only decrypted inside the run engine. Nothing sensitive is stored in plaintext, including your own API keys.
Sign up, take the starter token grant, and dispatch a run under full instrumentation within minutes.