Three Claude-powered tools are available to every team at Hydryx. Pick your tool, pick your team, and see exactly how it fits your day-to-day work.
Claude Code is a command-line AI tool (integrates with any IDE) that works across your repo and the files you give it access to and works directly alongside you to write, debug, refactor, and ship. It understands Hydryx's pipeline from MQTT broker through SQS and the data parser into the production database, so it gives answers grounded in your actual architecture, not generic examples.
Reads your entire repo — Django models, Lambda handlers, Terraform modules — not just a snippet.
Traces errors from Sentry stack traces down to root cause and produces a targeted patch with a diff.
Writes new features, API endpoints, ECS tasks, or Lambda handlers from a plain-English description.
Modernises legacy patterns, splits large functions, removes duplication — file by file, with diffs.
Writes unit and integration tests using your existing framework and fixtures, covering edge cases.
Writes optimised queries and pandas transforms against your real schema definitions in the codebase.
Scans pull requests for logic errors, security issues, and style problems before a human reviewer opens them.
Reviews Terraform plans, spots misconfigured IAM policies, and suggests cost or security improvements.
When the data parser throws an exception processing a message from SQS, paste the error into Claude Code. It reads the parser logic and SQS consumer code, identifies whether the issue is a malformed MQTT payload, a missing field, or a queue handling bug, and proposes a targeted fix — without needing to reproduce the issue locally.
Describe the endpoint in plain English and Claude Code writes the Django view, serializer, URL route, and tests — following Hydryx's existing patterns and conventions. It won't touch unrelated modules or introduce dependencies you don't need.
The Lambda algorithm fuses methane %, oxygen %, differential pressure, vacuum pump state, and KNMI weather data to compute a solenoid valve set-point per well. With so many interacting inputs, changing one rule can have unexpected downstream effects. Ask Claude Code to explain the current logic, safely add or tune a rule, and generate tests that cover all the edge cases across every metric combination.
Before running terraform apply, ask Claude Code to review the plan. It checks for overly permissive IAM roles, missing resource tags, and any changes that could affect the live data pipeline — SQS, the data parser, RDS, or the broker configuration for either Cedalo or AWS IoT Core.
terraform plan -out=plan.tfplan then show it to ClaudeDescribe the screen you need — a per-well gas-flow dashboard, a valve status table, an alarm history timeline — and Claude Code generates a React component that matches the Portal's existing patterns, state management, and API calls.
Ask Claude Code to write integration tests covering the full path: MQTT payload arrives in SQS, the data parser picks it up and processes it, and the record lands correctly in the production database. It includes edge cases like duplicate messages, out-of-range sensor values, and queue retries — using your existing fixtures.
Describe the business question in plain English. Claude Code reads your RDS schema definitions from the codebase, writes an optimised SQL query against the right tables, and explains every join and filter so you understand exactly what you're running — ready to execute on the read-only replica.
Ask Claude Code to write queries that surface anomalies directly from the production database — wells with sudden pressure drops, methane spikes, or oxygen sensor drift. It knows the table structure from the codebase so you don't have to reverse-engineer the schema before writing a single line of SQL.
New to the database? Ask Claude Code to explain what the schema looks like — which tables exist, what each field means, how wells, sites, sensor readings, and valve commands are stored. It reads the codebase (models, migrations, ORM definitions) to give you an accurate picture without needing to bother a developer.
Ask Claude Code to write the Python code for your notebook — connecting to the read-only RDS replica, running SQL via SQLAlchemy or psycopg2, loading results into a pandas DataFrame, and applying transforms like resampling, rolling averages, or outlier detection. It reads the schema from the codebase so it uses the correct column names.
The internal portal at hydryx.me pulls well and site data from the production database. Ask Claude Code to explain which queries back a given view, why a number looks different between the portal and your notebook, or how to add a new data point to an existing page.
Paste a slow query and Claude Code identifies the bottleneck — missing indexes, unnecessary full-table scans, poor join order — and rewrites it with an explanation of each change. Especially useful as the volume of sensor readings grows over time and ad-hoc reports start timing out.
When adding a new sensor, ask Claude Code to design the MQTT payload schema, validate it against what the data parser expects, and flag any fields that need a parser update before the device ships. It's also useful when testing whether a message will be handled the same way by both the Cedalo broker and AWS IoT Core during the parallel-running period.
The Lambda algorithm sends valve commands to Bray and EBRO actuators in the field. Ask Claude Code to explain the command format, write test cases for boundary values (fully open, fully closed, partial set-points), and verify the command serialisation matches what the hardware expects.
When a device switches from cellular to satellite fallback, messages may arrive out of order or delayed, causing the SQS FIFO buffer to build up. Ask Claude Code to trace the retry logic in the parser, identify message deduplication gaps, and suggest improvements to handle flaky connectivity gracefully.
Methane, oxygen, and differential pressure readings all have expected physical ranges. Ask Claude Code to write a validation layer in the Python parser that rejects or flags out-of-range values, logs them to Sentry, and prevents bad calibration data from polluting the Aurora operational database.
Device messages are end-to-end encrypted over MQTT before hitting the external broker and then AWS IoT Core. Ask Claude Code to audit the encryption configuration, check certificate handling in the comms module code, and flag any misconfigurations that could expose sensor data in transit.
Ask Claude Code to read the comms module and MQTT payload code and produce a clear protocol specification — message types, field names, units, valid ranges, and retry behaviour. Useful when onboarding a new hardware supplier or sharing the spec with the software team.
Before driving to a landfill site, ask Claude Code to write a query for all devices at that location — last-seen time, current valve state, recent alerts, and wells flagged as offline. Run it on the read-only replica, arrive informed, and discover problems before you're on-site.
Before changing a device's set-points or communication parameters on-site, ask Claude Code to explain what that configuration controls — how it feeds into the Lambda algorithm's actuator decisions, which sensor thresholds it affects, and what to watch for after applying the change.
Ask Claude Code to write a SQL query that scans all active devices and finds any that haven't sent a reading in the last 24 hours. Run it against the read-only replica and paste the results into a Slack message or email to the software team — no coding needed on your end.
After a site visit, paste your rough notes into Claude Code and ask it to format them into a clear bug report — what device, what site, what symptoms were observed, what was tried, and what the software team needs to investigate. No more emails that get lost or misunderstood.
When sales needs a customer health update, ask Claude Code to write a SQL query for that site's uptime, average methane capture, valve operation counts, and alert history. Run it against the read-only replica, then paste the results into a customer-facing update or hand them to sales.
Ops teams see the product from a perspective no one else does. When you notice a recurring friction point — a configuration step that's too manual, a missing alarm, a Portal view that doesn't show the right info on-site — ask Claude Code to help you write it up as a clear, actionable feature request for the software team.
If a prospect is already a Hydryx customer or in a trial, ask Claude Code to write a SQL query for their site — number of active wells, average methane capture, device uptime. Run it on the read-only replica and go into the call with real numbers, not vague claims.
When a prospect asks a technical question — "How does the algorithm decide when to open a valve?", "What happens if cellular goes down?", "How does data get to Blockbax?" — ask Claude Code to explain it in plain English using the actual codebase, so your answer is accurate, not approximate.
Before a renewal meeting, ask Claude Code to write a SQL query for a customer's methane capture improvement, device uptime, and alert history since go-live. Run it against the read-only replica, then turn those numbers into a clear value story that makes the renewal conversation easy.
Ask Claude Code to write a SQL query that flags customers with declining device uptime, increasing alert frequency, or wells that haven't sent data in several days. Run it on the read-only replica and get ahead of churn by reaching out before the customer notices the problem.
When preparing a proposal for a new landfill site, ask Claude Code to write a query pulling benchmarks from similar existing sites — average wells monitored, typical methane capture improvement, standard deployment timeline. Run it on the read-only replica and include the real figures in your proposal.
When customers tell you what's missing or frustrating — a missing Portal view, an alert they wish existed, a report they have to build manually — ask Claude Code to help structure that feedback into a clear product request with business context, so it gets properly evaluated rather than sitting in a notes doc.
Claude Cowork turns one-off prompts into reusable skills. Marketing can publish a blog draft in 2 minutes. Finance can reconcile invoices without copy-pasting. Ops can run a daily Slack briefing automatically. No prompting expertise required — just click and run.
Activate purpose-built workflows (blog writer, invoice hunter, PDF tools) with a single click.
Connects to Slack, Gmail, Webflow, Google Drive, and more — no glue code needed.
Encode multi-step processes so anyone on the team can run them consistently every time.
Share an AI session across teammates so everyone works from the same context and output.
Set agents to run on a schedule — daily reports, weekly digests, recurring data pulls.
Build skills tailored to Hydryx's internal processes and data sources.
When something breaks in the Hydryx pipeline, Cowork can turn raw logs, Slack messages, and engineer notes into a structured incident summary. Instead of every person documenting issues differently, the team gets the same format every time: what happened, likely impact, affected component, open questions, and next actions.
After a deployment or system change, Cowork can generate a clear internal update for Hydryx teams: what changed, what teams should know, whether any process changed, and what to watch in the portal or device data afterwards.
Software teams lose time when requests arrive as vague messages. A Cowork skill can convert rough requests into a clean engineering-ready format with context, priority, expected outcome, and the part of the Hydryx system likely involved.
When analysts already have SQL or notebook outputs, Cowork can turn those raw numbers into a consistent weekly summary for the wider business: site trends, methane performance, offline wells, notable anomalies, and key actions to investigate.
Cowork can help the data team handle incoming requests from ops, sales, or leadership by translating vague asks into a clean analysis brief: business question, suggested metrics, time period, entities involved, and expected output.
When a metric differs between hydryx.me, a spreadsheet, or a notebook output, Cowork can help create a short explanatory note for non-technical teams. This avoids repeated back-and-forth and makes reporting differences understandable across Hydryx.
Hardware teams often work from long supplier documents. Cowork can turn a Bray or EBRO valve document, sensor spec sheet, or connectivity manual into a short Hydryx-ready summary: key limits, required settings, compatibility notes, and field implications.
When the hardware team finds an issue during testing or deployment, Cowork can convert rough observations into a structured handover for software or ops: device context, expected behaviour, actual behaviour, suspected cause, and what should happen next.
Cowork can turn repeated explanations into reusable onboarding assets: how Hydryx devices communicate, common failure modes, what to check during device prep, and which supplier docs matter most.
Cowork is a strong fit for repeatable ops communication. A shared workflow can generate a daily internal briefing with key device issues, sites needing attention, offline units, and any follow-ups from the previous day.
Before a site visit, Cowork can compile a standard pre-visit pack: site notes, recent issues, known device problems, open actions, and customer context. This keeps every field visit prepared, even when someone new is covering the site.
After a visit, Cowork can take rough notes and convert them into a structured summary for internal teams or customers: what was checked, what changed, what issues remain, and whether follow-up is needed from software, hardware, or sales.
Sales teams repeat the same proposal work constantly. Cowork can turn a few inputs — customer type, site context, scope, and benchmarks — into a first-draft proposal structure with the right sections, tone, and Hydryx framing.
After prospect or customer calls, Cowork can convert rough notes into a consistent follow-up package: summary email, open questions, internal action items, and a short CRM-ready note.
When the same feedback keeps appearing across customers, Cowork can collect and standardise it for internal teams. This helps Hydryx spot repeated product requests, reporting needs, or customer confusion patterns earlier.
Claude Chat is the fastest way to get expert help with writing, research, contracts, and strategy — no technical knowledge needed. Just describe what you need as you would to a brilliant colleague, and get a thorough, thoughtful answer.
Thinks through complex problems step by step — strategy, trade-offs, and nuanced decisions.
Drafts, rewrites, and polishes emails, proposals, reports, and presentations.
Reads, summarises, and extracts key information from uploaded PDFs, contracts, and reports.
Aggregates information on any topic and delivers structured, actionable summaries.
Generates ideas, challenges assumptions, and stress-tests proposals with fresh perspectives.
Translates and adapts content across languages while preserving tone and intent.
Describe the situation — a customer asking about delayed installation, a partner inquiry about Blockbax data access, or an internal escalation — and Claude writes a clear, professional email in your tone.
Upload a contract from a valve supplier, a cloud vendor, or a new SaaS tool and ask Claude to summarise the key clauses, flag unusual terms (liability caps, data ownership, auto-renewals), and explain what you're actually agreeing to.
Ask Claude to research EU landfill gas regulations, methane emission standards, competitor monitoring approaches, or IoT connectivity trends. Receive a structured briefing with context and key takeaways — ready to share or present.
Share a product roadmap decision, a pricing proposal, or a customer pitch and ask Claude to play devil's advocate. Get a structured critique — risks, assumptions you haven't questioned, and alternatives — before you commit.