AI Tools Guide
Welcome to Hydryx

Your AI toolkit,
explained simply

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

Your AI pair programmer, right in the terminal

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.

$ claude
✦ Reading hydryx-parser repo…
✦ Found SQS timeout in mqtt_ingest.py
✦ Generating fix + test…
✔ Patch ready. Sentry alert resolved.
What's next?
🔍

Codebase understanding

Reads your entire repo — Django models, Lambda handlers, Terraform modules — not just a snippet.

🐛

Bug detection & fixes

Traces errors from Sentry stack traces down to root cause and produces a targeted patch with a diff.

✍️

Code generation

Writes new features, API endpoints, ECS tasks, or Lambda handlers from a plain-English description.

♻️

Refactoring

Modernises legacy patterns, splits large functions, removes duplication — file by file, with diffs.

🧪

Test generation

Writes unit and integration tests using your existing framework and fixtures, covering edge cases.

📊

SQL & data pipelines

Writes optimised queries and pandas transforms against your real schema definitions in the codebase.

🔎

PR & code review

Scans pull requests for logic errors, security issues, and style problems before a human reviewer opens them.

🏗️

Terraform & infra

Reviews Terraform plans, spots misconfigured IAM policies, and suggests cost or security improvements.

ℹ️ Full-stack pipeline work — Django REST, Python ECS parser, React Portal, Lambda algorithm, Terraform.
🐛
Debugging · Data Parser

Trace a parser failure back to the SQS message in minutes


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.

  • 1 Paste the error or stack trace: "What's causing this?"
  • 2 Claude reads the parser, SQS handler, and schema definitions
  • 3 Review the root-cause explanation and accept the patch
⚡ 1–3 hours of investigation → under 10 minutes
🚀
Feature Development · Django REST

Add a new Portal API endpoint without touching unrelated code


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.

  • 1 "Add a GET /wells/{id}/latest-reading endpoint"
  • 2 Claude reads existing views and models, writes the implementation
  • 3 Review the diff, run tests, and push — no boilerplate to write
⚡ First working draft in under 5 minutes
🤖
Actuator Control · Lambda Algorithm

Configure multi-sensor actuator logic without breaking other inputs


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.

  • 1 "What happens to the valve set-point when O₂ is high but CH₄ is also high?"
  • 2 Claude traces the three algorithm stages and shows exactly how all inputs combine
  • 3 Make a targeted change; Claude writes tests covering affected sensor combinations
⚡ Safe actuator tuning without unintended side effects
🏗️
Infrastructure · Terraform

Review Terraform changes before applying them to AWS


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.

  • 1 Run terraform plan -out=plan.tfplan then show it to Claude
  • 2 Claude flags security, cost, or availability risks
  • 3 Fix flagged issues before the apply — no surprises in prod
⚡ Catch infra issues before they become incidents
⚛️
Frontend · React Portal

Build new Portal views from a description, not a blank file


Describe 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.

  • 1 "Build a table showing all wells with last-seen methane reading and valve state"
  • 2 Claude reads existing components and writes a consistent new one
  • 3 Iterate on layout or data — no boilerplate or copy-paste
⚡ UI components in minutes, not hours
🧪
Testing · Full Pipeline

Generate integration tests for the MQTT → Aurora path


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.

  • 1 "Write integration tests for the data parser end-to-end flow"
  • 2 Claude reads the parser, DB models, and existing tests
  • 3 Run the suite; ask Claude to fix any failures inline
⚡ High-coverage test suite without writing it from scratch
📊
Production DB · SQL Queries

Write well-performance queries against the real schema


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.

  • 1 "Show average methane flow per well for the last 30 days, grouped by site"
  • 2 Claude reads the schema definitions and writes the SQL for you
  • 3 Run it on the read-only replica; iterate on filters or time windows
⚡ Complex queries in minutes — no schema guessing
🔍
Anomaly Detection · RDS

Write queries to surface sensor anomalies from the database


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.

  • 1 "Find wells where differential pressure dropped more than 20% in an hour"
  • 2 Claude writes the SQL query with the correct columns and joins
  • 3 Run it on the read-only replica and pipe results into a notebook
⚡ Surface anomalies without building a separate analytics layer
ℹ️ Claude writes the query for you to run — it doesn't connect to the database directly.
🗄️
RDS · Schema exploration

Understand the production database without asking the dev team


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.

  • 1 "Explain the well and site tables and how they relate to sensor readings"
  • 2 Claude reads the codebase and produces a clear schema summary
  • 3 Ask follow-up questions or generate a query straight away
⚡ Onboard to the schema in 20 minutes instead of a full day
🐍
Python Notebooks · RDS

Write notebook queries without guessing column names


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.

  • 1 "Write notebook code to pull 1-hour rolling avg methane per well for the last 7 days"
  • 2 Claude writes the connection + query + pandas transform with correct column names
  • 3 Paste into your notebook, run, and iterate
⚡ Notebook analysis in minutes, not half a day
🌐
Hydryx Portal · hydryx.me

Understand what data the portal displays and how to extend it


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.

  • 1 "What query does the well overview page use to calculate average flow?"
  • 2 Claude reads the portal code and shows you the exact SQL
  • 3 Cross-check with your own analysis or suggest a correction
⚡ No more guessing why portal numbers differ from your analysis
🐢
Performance · RDS Query Optimisation

Optimise slow queries on the read-only replica without a DBA


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.

  • 1 Paste the query and its EXPLAIN output
  • 2 Claude identifies the bottleneck and rewrites the query
  • 3 Compare runtime before and after — iterate if needed
⚡ Fast queries without waiting for a backend engineer
📡
MQTT · Payload Design

Design and validate MQTT message schemas for new sensors


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.

  • 1 "Design an MQTT payload for a new temperature sensor alongside the existing fields"
  • 2 Claude reads the parser schema and designs a compatible payload
  • 3 Get a list of parser changes needed before deployment
⚡ Catch parser/hardware mismatches before the device is in the field
🔌
Valve Commands · Bray / EBRO

Understand and test valve command formatting


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.

  • 1 "Show me how valve commands are structured and what the valid range is"
  • 2 Claude reads the Lambda send-command stage and explains the format
  • 3 Generate tests for edge values — and confirm they reach the device correctly
⚡ Validate command format without field testing every edge case
🛠️
Connectivity · Cellular / Satellite

Debug intermittent connectivity and SQS backlog issues


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.

  • 1 "Why are we seeing duplicate readings when a device reconnects after satellite fallback?"
  • 2 Claude reads the SQS consumer and deduplication logic
  • 3 Propose a fix and write a test simulating the reconnect scenario
⚡ Resolve connectivity edge cases before they affect data quality
🧪
Sensor Calibration · Data Validation

Write validation scripts for incoming sensor readings


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.

  • 1 "Add sensor range validation for CH4 (0–100%), O2 (0–25%), and dP (-500 to 500 Pa)"
  • 2 Claude writes the validation logic inside the ECS parser
  • 3 Review the Sentry integration and deploy — bad data never reaches Aurora
⚡ Protect data quality without writing boilerplate validation by hand
🔐
Security · MQTT Encryption

Audit the encrypted MQTT path from device to AWS IoT Core


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.

  • 1 "Audit our MQTT TLS configuration for the device comms module"
  • 2 Claude reviews cert pinning, cipher suites, and handshake logic
  • 3 Receive a ranked list of findings with recommended fixes
⚡ Security audit in minutes, not a dedicated pen-test sprint
📖
Documentation · Device Protocol

Auto-generate device communication docs from source code


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.

  • 1 "Generate a protocol spec from the device comms and parser code"
  • 2 Claude reads both sides and produces a structured specification
  • 3 Export as Markdown or PDF — share with any team or supplier
⚡ Protocol docs in 10 min instead of a 2-day write-up
🗺️
Site Preparation

Get a ready-to-run query for device status before a field visit


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.

  • 1 "Write a query for all devices at Site X — last reading, valve state, alerts"
  • 2 Claude writes the SQL; you run it on the read-only replica
  • 3 Use the results to prioritise which wells to check first
⚡ Arrive prepared — no surprises on-site
🔧
Device Configuration

Understand what a config change will do before applying it


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.

  • 1 "What does changing the minimum flow set-point do to the algorithm's valve decisions?"
  • 2 Claude reads the algorithm Lambda and explains the downstream effect
  • 3 Apply the change confidently, knowing what to monitor afterwards
⚡ No more config changes that cause unexpected behaviour
📋
Fleet Monitoring

Write a query to flag devices that haven't reported in 24 hours


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.

  • 1 "Write a query to find all devices with no reading in the last 24 hours"
  • 2 Claude writes the SQL with device ID, site, and last-seen timestamp
  • 3 Run it daily on the read-only replica as a quick morning health check
⚡ Catch silent device failures before customers do
📝
Cross-team Communication

Turn field notes into a structured bug report for the software team


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.

  • 1 Paste raw field notes: "Device at Well 4 keeps dropping MQTT connection after ~2hrs"
  • 2 Claude formats them into a structured issue with relevant context
  • 3 Send to the software team — they have everything they need to investigate
⚡ Field issues get fixed faster when they're clearly reported
📊
Customer Reporting

Generate a per-site performance query to share with sales


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.

  • 1 "Write a query for Site X performance over the last 90 days — uptime, methane flow, alerts"
  • 2 Claude writes the SQL; you run it and get a clean result set
  • 3 Hand the numbers to sales without needing to involve the data team
⚡ Customer updates without chasing the data team each time
💡
Product Feedback

Structure improvement suggestions into actionable feature requests


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.

  • 1 Describe the problem: "We always have to check two screens to find the last valve command"
  • 2 Claude formats it as a feature request with context, impact, and a suggested solution
  • 3 Submit it to the backlog — with enough detail that it actually gets prioritised
⚡ Field insights become product improvements, not lost Slack messages
🏭
Prospect Research

Get a site-level data query ready before a prospect call


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.

  • 1 "Write a query for Site X's performance over the last 60 days"
  • 2 Claude writes the SQL; you run it and get uptime, methane flow, and alert counts
  • 3 Use the numbers in your call — concrete data builds trust faster than slides
⚡ Data-backed conversations without chasing the ops team
Technical Questions

Understand how the product works without asking an engineer


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.

  • 1 "Explain how the algorithm decides valve set-points based on methane and oxygen readings"
  • 2 Claude reads the Lambda algorithm and returns a plain-English explanation
  • 3 Answer the prospect confidently — no need to loop in an engineer
⚡ Answer technical questions in the call, not 2 days later
📈
Customer Success

Generate a ROI summary for a renewal or upsell conversation


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.

  • 1 "Write a query comparing Site X's methane capture in Q1 vs. Q4 last year"
  • 2 Claude writes the SQL; you run it and get the comparison with figures
  • 3 Use it as the opening slide in the renewal meeting
⚡ Renewals backed by data close faster
🔔
Customer Health

Spot at-risk customers before they raise a complaint


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.

  • 1 "Write a query for customer sites with more than 3 device alerts this week"
  • 2 Claude writes the SQL; you run it and get a ranked list with context
  • 3 Proactively reach out — customers notice you care before they complain
⚡ Reduce churn by spotting issues before customers do
📄
Proposals

Generate a tailored technical summary for a new site proposal


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.

  • 1 "What's the average methane capture improvement across sites with 10–20 wells?"
  • 2 Claude writes the SQL; you run it across similar deployments to get benchmarks
  • 3 Include the figures in your proposal — specific beats generic every time
⚡ Proposals with real data win more often
🗣️
Feedback Loop

Translate customer feedback into structured input for the product team


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.

  • 1 Describe the customer's pain point in plain language
  • 2 Claude formats it as a structured request with impact and suggested solution
  • 3 Send it to the product backlog — customer voice, clearly articulated
⚡ Customer insights actually reach the roadmap
🤝 Claude Cowork

Shared AI workflows anyone on the team can trigger

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.

▸ Blog Writer skill active
Input: LinkedIn post URL
✦ Drafting article…
✦ Uploading to Webflow CMS
✔ Draft ready for review
Shared with: marketing@hydryx.io
🧩

One-click skills

Activate purpose-built workflows (blog writer, invoice hunter, PDF tools) with a single click.

🔗

Tool integrations

Connects to Slack, Gmail, Webflow, Google Drive, and more — no glue code needed.

🔄

Repeatable workflows

Encode multi-step processes so anyone on the team can run them consistently every time.

👥

Shared sessions

Share an AI session across teammates so everyone works from the same context and output.

Scheduled agents

Set agents to run on a schedule — daily reports, weekly digests, recurring data pulls.

🏗️

Custom skill creation

Build skills tailored to Hydryx's internal processes and data sources.

ℹ️ Reusable internal workflows for recurring software-team communication, documentation, and issue triage.
🚨
Incident Handling

Turn parser or pipeline incidents into a consistent triage report


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.

  • 1 Paste the incident notes, error messages, or Slack thread into the skill
  • 2 Claude turns them into a standard Hydryx incident report
  • 3 Share the output with the team or attach it to your internal tracker
⚡ Faster handovers and fewer half-documented incidents
📘
Internal Documentation

Turn technical changes into release notes and internal updates


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.

  • 1 Provide the change summary, PR notes, or rollout notes
  • 2 Claude drafts a clean internal release update
  • 3 Share it in Slack, email, or your internal docs
⚡ Keep non-engineers informed without writing every update from scratch
🧭
Cross-team Requests

Standardise incoming requests from ops, sales, and data


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.

  • 1 Drop in the raw request or message thread
  • 2 Claude extracts the key problem, business context, and missing details
  • 3 Output a standard request format for the backlog or team queue
⚡ Less time deciphering requests, more time solving them
📈
Recurring Reporting

Generate a weekly Hydryx performance summary from analyst outputs


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.

  • 1 Upload the CSV, spreadsheet, or notebook output
  • 2 Claude writes the summary in a standard Hydryx reporting format
  • 3 Send it to Slack, Drive, or email for the wider team
⚡ Analysis becomes shareable without manual write-up every week
🧹
Analysis Intake

Turn business questions into structured analyst briefs


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.

  • 1 Paste the original request into the skill
  • 2 Claude rewrites it as an analyst-ready brief
  • 3 Save or share the brief before starting the analysis
⚡ Cleaner analysis requests, especially useful for new analysts
🗂️
Portal / Reporting Alignment

Document why portal numbers and analysis outputs differ


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.

  • 1 Provide the two outputs and a short explanation of the difference
  • 2 Claude drafts a business-friendly explanation
  • 3 Reuse it in Slack, docs, or a report appendix
⚡ Fewer repeated questions about “why the numbers don’t match”
📄
Supplier Documentation

Summarise valve, sensor, and comms supplier PDFs into usable notes


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.

  • 1 Upload the supplier PDF or manual
  • 2 Claude extracts the most relevant technical details
  • 3 Save the output as a reusable internal reference
⚡ Turn long vendor docs into practical field-ready notes
🛠️
Field-to-Engineering Handover

Turn hardware test notes into a structured issue or change request


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.

  • 1 Paste the raw notes, photos, or observations into the skill
  • 2 Claude organises them into a clear issue summary
  • 3 Share the result with the right internal team
⚡ Better communication between field, hardware, and software
📚
Onboarding

Create internal setup and troubleshooting guides for new hardware teammates


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.

  • 1 Collect existing notes, checklists, and source docs
  • 2 Claude combines them into a standard guide
  • 3 Reuse it for every new teammate or supplier handover
⚡ Reusable know-how instead of repeating the same explanations
🌅
Daily Operations

Run a daily fleet briefing for the ops team


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.

  • 1 Connect the inputs once: notes, exports, or internal sources
  • 2 Claude compiles the update into a standard daily format
  • 3 Send it automatically to Slack or email each morning
⚡ Replace the recurring manual morning status message
🚗
Field Visits

Generate a site-visit briefing pack before leaving for the field


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.

  • 1 Select the site or customer
  • 2 Claude pulls together the relevant notes and context
  • 3 Use the briefing on-site or share it with the rest of the team
⚡ Better-prepared site visits with less last-minute digging
📝
Post-Visit Reporting

Turn raw field notes into a clean customer or internal update


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.

  • 1 Paste your raw visit notes into the skill
  • 2 Claude writes a clear, shareable summary
  • 3 Send it internally or adapt it for the customer
⚡ Faster handovers after every field visit
📦
Proposal Prep

Generate a first-draft proposal pack using Hydryx templates


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.

  • 1 Enter the prospect details and proposal context
  • 2 Claude drafts the proposal in the Hydryx format
  • 3 Review and personalise instead of starting from zero
⚡ First draft in minutes instead of rebuilding every proposal manually
🤝
Meeting Follow-up

Turn call notes into structured follow-up emails and CRM-ready summaries


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.

  • 1 Paste in the meeting notes or transcript
  • 2 Claude creates the follow-up email and internal summary
  • 3 Send externally and save the internal version for the team
⚡ Better follow-up quality, especially after busy call days
🔄
Customer Feedback Loop

Convert recurring customer feedback into a reusable internal summary


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.

  • 1 Add call notes, emails, or feedback snippets
  • 2 Claude groups them into themes and rewrites them clearly
  • 3 Share the output with product, ops, or leadership
⚡ Repeated customer signals become visible faster
💬 Claude Chat

Expert-level thinking, for every role at Hydryx

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.

You: Summarise this supplier contract
Claude: Key terms —
• Duration: 24 months, auto-renew
• Liability cap: €50 000
• Delivery SLA: 10 working days
You: Any risks?
🧠

Deep reasoning

Thinks through complex problems step by step — strategy, trade-offs, and nuanced decisions.

✍️

Writing & editing

Drafts, rewrites, and polishes emails, proposals, reports, and presentations.

📑

Document analysis

Reads, summarises, and extracts key information from uploaded PDFs, contracts, and reports.

🔬

Research synthesis

Aggregates information on any topic and delivers structured, actionable summaries.

💡

Brainstorming

Generates ideas, challenges assumptions, and stress-tests proposals with fresh perspectives.

🌍

Translation & localisation

Translates and adapts content across languages while preserving tone and intent.

📧
Communication

Draft professional client or partner emails in seconds


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.

  • 1 Describe who you're writing to and the key message
  • 2 Claude drafts a complete, polished email
  • 3 Tweak the tone or detail until it's right, then send
⚡ 2-minute draft vs. 20 minutes staring at a blank page
📜
Legal & Contracts

Understand hardware supplier or SaaS contracts without a lawyer


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.

  • 1 Upload the PDF contract
  • 2 "Summarise and flag anything risky"
  • 3 Use the summary to guide your negotiation or sign-off
⚡ First-pass review in under 3 minutes
📋
Research

Get a briefing on any landfill gas regulation or market topic


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.

  • 1 "Summarise EU landfill gas monitoring regulations for 2025"
  • 2 Claude returns a structured briefing with key points
  • 3 Ask follow-up questions to drill into specific areas
⚡ Research briefing in minutes, not half a day
🎯
Strategy

Stress-test a product decision or business proposal


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.

  • 1 Describe the decision or paste the proposal
  • 2 "Find the weaknesses and what I might be missing"
  • 3 Refine your thinking based on the critique before presenting
⚡ Better decisions before they're made public

Ready to get started?

Your Hydryx account gives you access to all three tools. Reach out to IT or your team lead to get set up.