One agentic run, from your records to a decision you can sign
A rule check, a plan you approve, then six cited sections from your own records
the pilot
What a pilot runs
Three stages, from your records to a document your engineers have scored.
Well records and brine chemistry
Your records in
You load well records, brine analyses and completion reports, and a rule check runs first.
The corrosion and materials basis of design
Cited sections out
Agents write the six sections, cite each claim to its page and name every gap.
Open the platformEvaluation inside the pilot
Scored with you
Your engineers score each run against fixed questions, and we fix every miss.
the run
You approve the plan before anything is written
Agents propose six sections, you approve the plan, and each section builds on the last.
Before the first section
Intake and the gate
The intake
Geothermal production well casing, SS 316 / 316L, high TDS brine with CO2 and H2S, 500 psi, 25 year design life
- 180 C downhole
- chloride 45,000 mg/L
- pH 5.8 at temperature
- H2S 0.2 bar
- CO2 1.5 bar
- single spot sample
the policy gate
Some questions must not get a fluent answer
A fixed rule check runs before any answer and refuses what the record cannot support.
Policy decision
Declined, on purpose
On some questions a fluent answer is the wrong product. This one was decided before retrieval, and what would change it is listed below.
No qualified materials envelope exists for supercritical or superhot geothermal service.
Materials qualification for geothermal service rests on test data, and above roughly 375 C that data does not exist. NACE MR0175 and ISO 15156 are scoped to H2S containing oil and gas production and stop applying well below these temperatures. Autoclave testing, the most capable method in general use, reaches about 350 C. The honest position is that nobody knows: a specific alloy recommendation at these conditions would be an extrapolation dressed as an answer, and a well designed on it would find out during commissioning.
What would close the gap
- Autoclave or flowing loop test data at the actual service temperature, pressure and fluid chemistry, for the specific product form and weld condition
- Published field experience from a well that has run these conditions, with exposure duration and post service inspection
- A qualification programme scoped to the deviation: candidate alloys, the test matrix, and the acceptance criteria agreed with the engineer of record
- Until that exists, the design decision belongs to a qualification programme rather than a materials lookup
This is a refusal from the product's policy layer, not a retrieval failure. Nothing in the corpus would change it.
supercritical-materials-recommendation- Materials for supercritical wells, where no alloy has test data.
- Cost figures built on too few data points.
- Signing off a well as fit for service.
A fluent wrong answer on a well design is worse than no answer.
evidence
Every source shows how much review it has had
Each citation carries a rank from T1 to T4 for how much review its source has had.
- T1 Standard or measured field data
- Standards and field measurements that have passed formal review.
- T2 Peer reviewed literature
- Papers that other experts reviewed before they were published.
- T3 SME synthesis
- Summaries written by a subject expert for this library.
- T4 Conference or gray literature
- Conference papers and reports that had the least outside review.
The same wall, and the corrosion problem on opposite faces of it.
- EGSEnhanced geothermal system
The reservoir fluid is inside the casing
Attacked surface: the inside of the casing
- Flashing brine, silica, chloride and H2S
- Periodic mechanical descaling strips a coating
- No way to inspect or repair it downhole
The question it asks
Can a coating survive the service?
Why descaling settles itHide the full envelope reason
Descaling operations destroy coatings. Any well with a known scaling history will need reaming or jetting, which removes a downhole barrier coating with no way to inspect or repair the damage. Redirect to CRA or clad, or to scale management at the design stage.
- AGSAdvanced geothermal system, closed loop
The working fluid is sealed inside the loop
Attacked surface: the outside of the casing
- Formation brine through thermally cycled cement
- Uninspectable and unrepairable for the asset life
- The well is a heat exchanger, so a coating costs output
The question it asks
What is the consequence if it does not?
Why consequence is the questionHide the full envelope reason
In a closed loop the internal environment is a controlled working fluid, often benign, so the corrosion problem migrates to the outside of the casing, in contact with formation brine through cement that has been thermally cycled. That external surface is uninspectable and unrepairable for the asset life, so the question is not whether the coating survives but what the consequence of it not surviving is, given no intervention path. Because the well is a heat exchanger, any coating adds thermal resistance, and the heat transfer penalty must be flagged alongside corrosion performance. If the working fluid is sCO2, elastomer explosive decompression and dry versus wet CO2 chemistry become the dominant materials issues and most coating literature is irrelevant.
So the agent establishes the architecture before it answers a coating question at all, and an answer that does not name one is answering a question nobody asked.
Plot of chloride against temperature for 316L, UNS S31603. A bordered box marks the region where test data exists: temperature up to 150 degrees Celsius, chloride 175 to 166,000 milligrams per litre. Everything to the right of 150 degrees is hatched and carries no test data. The EGS operating window runs 150 to 250 degrees Celsius, all of it at or above that ceiling. A marker sits at 200 degrees Celsius and 166,000 milligrams per litre, 50 degrees past the tested ceiling at a chloride the tests do cover. The hatched region is drawn neutral, not as a failure: it marks absent data.
- Temperature: up to 150 °C
- Chloride: 175 to 166,000 mg/L
- pH: 5 to 8.1
- Qualification tier: lab tested
The box is not a safety rating, and the blank is not a failure. A well 50 °C past the edge is a well nobody has measured, so the agent says that, states how far past the bound you are, and names the test that would close it.
Horizontal bar chart of temperature ceilings on a scale from 0 to 450 degrees Celsius. Shaded bands mark the EGS operating window from 150 to 250 degrees and the superhot band from 375 to 450 degrees. Coatings: fusion bonded epoxy 120 degrees, novolac epoxy 175 degrees, PEEK or PTFE lined pipe 250 degrees, thermal spray aluminum 200 degrees, CRA weld overlay or clad 350 degrees. Inspection tools: standard e line logging tools 175 degrees, high temperature logging tools 260 degrees, UT couplant 200 degrees. Test methods: NACE TM0177 SSC testing 24 degrees, NACE TM0198 200 degrees, autoclave testing 350 degrees. Almost every ceiling sits below the geothermal operating window.
“What coatings survive geothermal” is four different questions. The agent pins down which one is being asked before it answers.
| What they mean | Real answer shape |
|---|---|
| Downhole production casing | Decline Essentially nothing is qualified. Show the full envelope reasonHide the full envelope reasonIn EGS the reservoir fluid is inside the casing, so the coating faces flashing brine, silica supersaturation, chloride, H2S, and periodic mechanical descaling. Reaming or jetting to remove scale destroys a barrier coating, and there is no way to inspect or repair it downhole. Coatings fail non gracefully: perfect until a holiday, then under film attack runs fast at 200 C, while a CRA thins predictably. For 200 C class EGS production casing the correct redirect is CRA or clad. |
| Downhole injection below 150 °C | Conditional Real candidates exist (GRE liner, internal epoxy) with caveats. Show the full envelope reasonHide the full envelope reasonConnection areas and holidays remain bare, handling and running damage over thousands of feet cannot be inspected after installation, depressurization can collapse or blister the lining, and coated carbon steel defeats most wall loss NDE. Viable only where the well has no scaling history and no mechanical descaling is anticipated. |
| Surface two phase piping and vessels | Recommended class External CUI systems, TSA, inorganic zinc and silicone. Show the full envelope reasonHide the full envelope reasonThis is where coatings genuinely work. Surfaces are inspectable and repairable, which removes the core downhole objection. Watch erosion in two phase flow with entrained silica and blistering on depressurization of flashing service. |
| Turbine, valve trim, heat exchanger | Erosion class HVOF cermet, diffusion aluminide, cladding. Show the full envelope reasonHide the full envelope reasonThe problem is erosion and H2S, not barrier corrosion. Field data supports thermal spray with a bond coat to 210 C. Components are accessible for inspection and periodic refurbishment. |
Downhole production casing
DeclineEssentially nothing is qualified.
Show the full envelope reasonHide the full envelope reason
In EGS the reservoir fluid is inside the casing, so the coating faces flashing brine, silica supersaturation, chloride, H2S, and periodic mechanical descaling. Reaming or jetting to remove scale destroys a barrier coating, and there is no way to inspect or repair it downhole. Coatings fail non gracefully: perfect until a holiday, then under film attack runs fast at 200 C, while a CRA thins predictably. For 200 C class EGS production casing the correct redirect is CRA or clad.
Downhole injection below 150 °C
ConditionalReal candidates exist (GRE liner, internal epoxy) with caveats.
Show the full envelope reasonHide the full envelope reason
Connection areas and holidays remain bare, handling and running damage over thousands of feet cannot be inspected after installation, depressurization can collapse or blister the lining, and coated carbon steel defeats most wall loss NDE. Viable only where the well has no scaling history and no mechanical descaling is anticipated.
Surface two phase piping and vessels
Recommended classExternal CUI systems, TSA, inorganic zinc and silicone.
Show the full envelope reasonHide the full envelope reason
This is where coatings genuinely work. Surfaces are inspectable and repairable, which removes the core downhole objection. Watch erosion in two phase flow with entrained silica and blistering on depressurization of flashing service.
Turbine, valve trim, heat exchanger
Erosion classHVOF cermet, diffusion aluminide, cladding.
Show the full envelope reasonHide the full envelope reason
The problem is erosion and H2S, not barrier corrosion. Field data supports thermal spray with a bond coat to 210 C. Components are accessible for inspection and periodic refurbishment.