Documentation

# Homebrew (recommended)
brew install ankit373/hydra/hyctl

# npm  ·  or run once, no install:  npx hyctl
npm install -g hyctl

# pip
pip install hyctl

# Standalone installer
curl -fsSL https://raw.githubusercontent.com/ankit373/hydra/main/install.sh | sh

# From source (Go 1.26+)
git clone https://github.com/ankit373/hydra.git && cd hydra
go build -o hyctl ./cmd/hydra

Chat executes: the route line names the routing level, model and strategy the router picked and why, then the proof strip reports what the run planned, edited and verified. shift+tab cycles the modes, Auto (plan → edit → run your tests → fix → repeat), Plan, Edit, Ask, and ctrl+o overrides where the next task runs without changing what it does. Models is the discovered estate by provider, each model at its routing level, with a server's children greyed when the server itself is down. Audit carries the calibration scorecard, the hash-chained log's integrity, the guardrails actually in force, and anything waiting on a human. Agents, Activity and Usage cover work in flight, per-run traces, and where the money went. ? lists every key, rendered from the keymap itself so it cannot drift.

An answer appears as it arrives rather than all at once when the task finishes, so a slow model and a stuck one no longer look the same. The frame repaints five times a second, which is also what coalesces the stream: deltas land every ~20ms and the render never runs per delta.

When a model gives up part-way through, its partial output collapses to one dim line naming the model and the reason, so it cannot be mistaken for the answer, and e expands it again. The text is kept, not re-fetched. Only plain routing streams: a swarm has several models answering at once and consensus picks by judge after N replies, so neither has a single stream to show.

hyctl tui
# tab cycles views · shift+tab modes · ctrl+o routing · e expands a
# collapsed partial · ? all keys · :q quit
hyctl tui --snapshot --view 0..5   # render one static frame

# Download from the releases page:
hydra-desktop_<version>_darwin_universal.zip   # macOS, Intel + Apple Silicon
hydra-desktop_<version>_windows_amd64.zip      # Windows
hydra-desktop_<version>_linux_amd64.tar.gz     # Linux

# Every artifact ships a .sha256 beside it. Verify before you run:
shasum -a 256 -c hydra-desktop_<version>_darwin_universal.zip.sha256   # macOS
sha256sum   -c hydra-desktop_<version>_linux_amd64.tar.gz.sha256       # Linux

# Windows (PowerShell):
$f = "hydra-desktop_<version>_windows_amd64.zip"
(Get-FileHash $f -Algorithm SHA256).Hash -eq (Get-Content "$f.sha256").Split(' ')[0]

# Or build it yourself:
go install github.com/wailsapp/wails/v2/cmd/wails@latest
cd desktop && wails build                # → desktop/build/bin/Hydra.app

export ANTHROPIC_API_KEY=sk-…      # then:
hyctl probe                        # it shows up as a scored model

A key alone gives one routing choice: a single model. Hydra already prices and scores OpenRouter's whole catalogue, so naming models turns each into its own routing choice the router chooses between. Admission is explicit: enumerating hundreds would bury this list. Naming nothing changes nothing.

# ~/.hydra/config.toml
[openrouter]
models = ["anthropic/claude-sonnet-4.5", "google/gemini-2.5-pro"]

hyctl probe then lists one entry per model and says how much of the catalogue is enabled ("2 of 423"). A name the catalogue does not hold is shown as not routable with the reason, never dropped.

hyctl init

hyctl probe

Model list comes from live discovery, grouped by the routing level each model would route at, so it cannot drift from the router; anything discovery found but nothing can drive is counted, not listed. The governor models context usage as a drifting process and estimates the first-passage probability of crossing the ceiling before the task ends, escalating before a threshold is hit.

hyctl status

1 · classify → enum/routing level · 2 · policy gate (PII can block before any network call) · 3 · select highest-CapScore available model at/under routing level · 4 · execute · 5 · on failure fall down the chain to local Qwen · 6 · log tokens + cost. ~1.1µs of decision overhead.

hyctl dispatch "refactor auth to JWT refresh"
hyctl dispatch --enum SIMPLE "write a User DTO"
hyctl dispatch --dry-run "add pagination"
hyctl dispatch --local "write unit tests"

On a terminal the answer is drawn as it arrives: a spinner covers the wait before the first token, then tokens append, then the token counts, elapsed time and measured TTFT print as a footer, since none of those numbers exist until the call returns. Measured on a local 7B: 435 deltas spread over 9.9s of a 10.0s call, where before there was nothing on screen for the whole ten seconds.

Piped output never streams and is byte-identical to before, so scripts are unaffected; --no-stream asks for the same buffered rendering on a terminal. When a model fails part-way through, its partial is collapsed off the screen and replaced by one line naming the model and the reason, because a partial left on screen reads as part of the answer. Nothing is erased once it has scrolled past the top of the window.

A fan-out has no single stream to show, so --swarm and --confidence draw a live panel instead: every model that was selected, with its state, elapsed time, tokens and the cost as it accrues. A confidence run adds the running log-odds against the threshold it has to cross, so an ensemble that stops after two models says why it stopped. The panel is erased when the run ends, and the result printed after it is unchanged.

hyctl dispatch --swarm --swarm-mode best "implement a rate limiter"
  --swarm-max-heads 5  --swarm-max-cost 0.05  --swarm-judge-tier 1

An SPRT accumulates the calibrated log-likelihood ratio across votes and stops at the boundary implied by your target. --file raises the bar by blast radius.

hyctl dispatch --confidence 0.95 "is this migration safe for prod?"

hyctl edit --file "$PWD/internal/foo.go" --enum MODERATE --prompt "add a nil check"

hyctl review --file internal/foo.go

hyctl parallel --tasks tasks.txt

A Beta-Bernoulli posterior over sensitivity/specificity → a per-vote LLR and a diagnostic power D (KL divergence, nats). A coin-flip source has D≈0 and is ignored. The defect-cost model prices a wrong answer and sets the required confidence.

hyctl trust calibration      # per-source se / sp / D
hyctl trust record --source ollama/qwen3:4b --domain go --said-correct --outcome correct  # a model ID from `hyctl probe`
hyctl trust defect --pii --production
hyctl trust stats   # samples saved vs fixed-N
hyctl trust explain <hash>   # the LLR ledger for a run
hyctl trust benchmark        # measured SPRT numbers

From graph.json it counts transitive dependents and computes κ = ⟨k²⟩/⟨k⟩ (Molloy-Reed). κ ≥ 2 ⟹ cascade-capable core ⟹ higher confidence bar.

hyctl graph blast internal/auth/token.go

Amdahl + coordination cost: minimizing T(n)=s+(1−s)/n+kn gives n*=√((1−s)/k). Independent files → ~6; coupled → ~2.

hyctl graph parallel internal/a.go internal/b.go

ρ = |gzip(C)|/|C| proxies the entropy rate; useful = L·ρ. Compact on falling ρ, not raw length.

hyctl context entropy internal/foo.go
cat notes.md | hyctl context entropy -

hyctl mcp check <tool> --agent A --resource R --action write
hyctl mcp log --denied     # what got blocked
hyctl mcp report          # allowed / denied by agent & tool

hyctl mcp registry sync             # pull the official MCP registry into a local cache
hyctl mcp registry scan             # list servers installed across Claude Code/Desktop, Cursor, Windsurf, VS Code
hyctl mcp registry audit            # resolve + score installed servers, update lifecycle state
hyctl mcp registry export --out DIR  # static index.html/index.json of audited servers
hyctl mcp registry backtest         # validate scoring against known real incidents (postmark-mcp, CVE-2025-6514)
hyctl mcp registry list             # audited servers by safety score
hyctl mcp registry clear <server>   # recover a server quarantined in error

hyctl oracle verify go test ./... --source verifier:go-test

hyctl cost

hyctl stats

hyctl pricing list [filter] [--json]
hyctl pricing refresh

hyctl models list
hyctl models add kimi-k3 --provider moonshot --cap-score 85
hyctl models sync   # import the OpenRouter catalog

SSNs, credit cards, emails, API keys/tokens, IPs, private-key material → re-routed to a local model.