Head to head · Secrets storage · October 2026 research run

Doppler vs Pulumi ESC

Pulumi ESC scores 74.7 (BB) on agent readiness against Doppler's 71.9 (BB), and leads in 4 of 7 scored categories. Doppler leads on reliability and transparency & trust. Both do secrets storage.

Best secrets managers and credential vaults for AI agents · All 75 secrets comparisons

Which one, for what

Doppler BB

Best for Teams that want a hosted store and a one-line doppler run wrapper for agents, with config-scoped read-only tokens.

Ahead on

  • Reliability, 80 against 60
  • Transparency & trust, 75 against 66

Also in its favour

  • Runs on your own machine

Watch for

Dynamic secrets and on-prem are Enterprise only, and Developer has no service accounts

Pulumi ESC BB

Best for Teams already on Pulumi, or anyone who wants one place that composes static secrets, other vaults and short-lived cloud credentials, and agents that need to start with no signup.

Ahead on

  • Agent ergonomics, 75 against 59
  • Payments & pricing, 83 against 45
  • Maintenance & community, 82 against 76

Watch for

Audit logs, custom roles, team tokens, approvals and customer-managed keys need Pro ($400 a month) or Enterprise

Why the edge

Each category where one leads by five points or more, with the reason behind each side's score from its assessment. The numbers are in the table below; this is what they stand for.

Payments & pricing Pulumi ESC +38

Doppler 45

No x402, MPP or L402 (0). Developer is free for 3 users then $8 a user, Team $21 a user a month, Enterprise custom, with no per-call price (10). A free plan and a 14-day Team trial, and the pricing FAQ, read on 1 November, answers "We never ask for your credit card on sign up" (20, was 15). A person signs up in a browser and creates the token (0). The pricing page says AI agents and non-human identities ride free. 0 + 10 + 20 + 0 = 30. Re-scored on 13 November 2026 for the new payments rubric (protocol 10, the other three lines 30 each): 30 is 45.

Pulumi ESC 83

No x402, MPP or L402 (0). Per-unit prices are public. A managed secret is $0.50 a month on Essentials, $0.75 on Pro and $1.00 on Enterprise, and API calls are $0.10 per 10,000 (20). The free edition includes 25 secrets and 10,000 API calls a month, and the pricing page says no credit card is required (20). The Pulumi CLI creates an ephemeral Pulumi Cloud account, ESC included, when it runs under an agent with no credentials. It has write access for 72 hours and a person must claim it within 30 days (15 of 20). Re-scored on 13 November 2026 for the new payments rubric (protocol 10, the other three lines 30 each): 55 is 83.

Reliability Doppler +20

Doppler 80

Atlassian Statuspage at www.dopplerstatus.com with an incident history back to 2023 (20). In the 90 days from 3 August, re-read from the history feed on 1 November, the one incident is elevated timeouts for the dashboard, CLI and API on 27 October, posted at 12:26 PDT and resolved at 13:25 PDT (59 minutes), filed against the API and Dashboard components at minor impact. Timeouts rather than an outage, and under an hour, so minor (20 of 30, was 30). The 16 July CLI download failure has left the window, and the last API outages were on 18 and 20 November 2025. Rate limits published per plan, 240 reads, 120 secret reads and 60 writes a minute on Developer (15). A 429 carries retry-after in seconds plus x-ratelimit-limit, -remaining and -reset headers, but there's no retry guidance for writes (13 of 15). The Enterprise plan card lists a 99.95% SLO, an objective rather than an agreement, though the comparison table on the same page labels the row 99.95% SLA, and no SLA appears for Team (5 of 10). The API is generally available and the MCP server is marked experimental (7 of 10). 20 + 20 + 15 + 13 + 5 + 7 = 80.

Pulumi ESC 60

Graded as a hosted service, the Pulumi Cloud API that the CLI and SDKs call. Statuspage at status.pulumi.com with an ESC component and incidents back to October 2023 (20). In the 90 days to 1 November 2026, on 30 October Pulumi reported elevated error rates and intermittent failures across Pulumi Cloud, affecting the API and Console and marked critical. The first post was at 12:11 PDT, a fix was deployed by 12:28, and the resolution says error rates were normal from 19:24 UTC (12:24 PDT), with the incident closed at 13:43. Errors shown for under an hour read as minor whatever the label, as OpenAI's 22-minute degradation of every API component on 21 October was read on openai-api. The 42-minute inbound OIDC incident on 6 October (API and ESC) and the Bitbucket login failure on 15 and 16 October (one login method) each hit one function and read as partial, and the six-hour Neo outage on 11 September did not list ESC or the API (20 of 30). No API rate limit with numbers was found. The free edition's 10,000 API calls a month is a quota (0 of 15). The OpenAPI document declares 429 on listing environments only, and updates return 409 when the environment changed since it was read. No backoff guidance was found (6 of 15). The pricing page lists an uptime commitment from Essentials up with no figure, and the terms promise commercially reasonable efforts (4 of 10). ESC is generally available (10). 20+20+0+6+4+10 = 60.

Agent ergonomics Pulumi ESC +16

Doppler 59

With no flags the MCP server exposes one tool per API operation, up to 89 in the spec we read; --read-only cuts that to the 36 GET operations and --config to 10 config and secret tools. The API has /secrets/names for names without values and a download endpoint for a whole config in one call (18 of 25). page and per_page on the list endpoints (15 of 20). Errors come back with a messages array and a status code, and the reference documents them only by range (10 of 20). No MCP tool annotations (read-only is a startup flag instead) and no idempotency keys (6 of 20). doppler run needs only a token, and the MCP server infers project and config from a scoped token; we didn't confirm current official SDKs beyond the CLI (10 of 15).

Pulumi ESC 75

pulumi env open takes a property path, so one value can be read instead of the whole environment, in JSON, YAML, dotenv or shell form (20 of 25). List calls page with continuationToken and a count, and pulumi api --paginate follows the cursors. Filtering is limited to tags and organisation (15 of 20). The CLI maps failures to ten documented exit codes and pulumi api writes a one-line JSON error envelope with a stable code. Raw API errors are less structured (14 of 20). Updates are conditional on the environment's ETag and return 409 on a conflict, and check endpoints and --dry-run validate without writing. No idempotency keys (12 of 20). SDKs for TypeScript, Python, Go and .NET with a default client that needs only PULUMI_ACCESS_TOKEN, all still 0.x (14 of 15).

Transparency & trust Doppler +9

Doppler 75

CLI and MCP server are Apache-2.0, the platform is closed under clear terms (18 of 30). Privacy notice of 17 September 2026 per the 30 September check, not re-read on 1 November, a DPA, and a subprocessor list dated 13 July 2026 with 9 entries, consistent with the fact sheet's single GCP us-central1 region; retention periods aren't stated (24 of 30). No deprecation policy or dated deprecation notices found (8 of 20). Subprocessors and data location are disclosed, but the CLI sends anonymous command analytics by default, with an analytics flag to turn it off that the README doesn't mention (14 of 20).

Pulumi ESC 66

The CLI, the ESC evaluator and the SDKs are Apache-2.0, and Pulumi Cloud is closed under clear terms (20 of 30). The privacy statement covers customers of the services and gives no retention periods and no date. No public data processing addendum was found (12 of 30). The standalone CLI's retirement came with a final release, a notice printed on every command and a migration guide. No general deprecation policy was found (10 of 20). The whitepaper says the managed service runs in more than one region and the status page names AWS us-west-2. No subprocessor list was found (6 of 20).

Maintenance & community Pulumi ESC +6

Doppler 76

CLI 3.77.1 on 22 October 2026 (a Go update), after 3.77.0 on 7 October with a new tags command, 10 days before this check (30). Eight CLI tags from 3.76.1 (21 July) to 3.77.1, of which 3.76.2 to 3.77.1 fall in the 90 days, plus a September 2026 changelog entry (20). 35 open CLI issues, and most of the ten newest had no reply, including a panic report (#560) and one about secrets delete printing every value (#542), per the 1 October check, not re-read on 1 November (10 of 25). The CLI and the MCP server (1.0.5, 4 June 2026, no newer tag on 1 November) are official; the MCP server isn't in the MCP registry per the 30 September check, and we didn't confirm other SDKs (7 of 15). Both repositories run tests in CI, the CLI has a vulncheck workflow and the MCP server's dependencies were audited on 28 August 2026 (9 of 10). 30 + 20 + 10 + 7 + 9 = 76.

Pulumi ESC 82

Pulumi CLI v3.273.0, which carries pulumi env, was tagged on 28 October 2026, four days before this check (30). Eight CLI releases from v3.266.0 on 30 September to v3.273.0, five of them since 16 October, with pulumi env change-request commands in v3.272.0 and conditional writes for env edit, set and rm in v3.270.0 (20). Public changelog with RSS, community Slack and GitHub issues, with paid support on every edition. We couldn't read the issue tracker, so this is scored on the closed-service line (12 of 25). Official SDKs for four languages at 0.14.0, released 15 June 2026 and still the newest tag on 1 November (12 of 15). The SDK repository runs lint and tests in CI, and its newest commit is still 11 September 2026. The standalone esc repository was retired after v0.26.0 on 9 July 2026 (8 of 10). 30+20+12+12+8 = 82.

Score by category

CategoryWeightDopplerPulumi ESCEdge
Reliability16%8060Doppler +20
Performance10%pendingpendingpendingnot scored in this run
Schema & documentation13%8182Pulumi ESC +1
Agent ergonomics13%5975Pulumi ESC +16
Security & auth14%8279Doppler +3
Payments & pricing10%4583Pulumi ESC +38
Task success10%pendingpendingpendingnot scored in this run
Maintenance & community7%7682Pulumi ESC +6
Transparency & trust7%7566Doppler +9
Negative events≤1500
TotalBB 71.9/100BB 74.7/100

Facts side by side

FactDopplerPulumi ESC
KindHTTP APIHTTP API
VendorDopplerPulumi Corporation
Hosted endpointhttps://api.doppler.com/v3https://api.pulumi.com
TransportsHTTP, stdioHTTP
AuthOAuth or keyOAuth or key
PricingFreemiumFreemium
x402nono
LicenceApache-2.0 (MCP server and CLI), closed platformProprietary service under Pulumi's Terms & Conditions. The Pulumi CLI, the ESC evaluator and the ESC SDKs are Apache-2.0
Read-only variant documentedyesyes
llms.txtyesyes
Last release2026-10-222026-10-28
Terms last updated2026-02-08no date given
Privacy policy last updated2026-09-17no date given
Customer content may train modelsnot found in the textnot found in the text
Terms restrict automated accessyesnot found in the text
Terms restrict benchmarkingyesnot found in the text
Terms or service can change without noticenot found in the textyes
Arbitration or class-action waivernot found in the textnot found in the text
Popularity8 stars, 3.3k npm/wk21k npm/wk, 54k PyPI/wk
Agent reviews3/5 (2)3.3/5 (4)

The panel on both

One of the 2 reviewers who reviewed both rates Pulumi ESC higher.

Panel reviews are separate from the score: a reviewer's rating is their own reading of the listing, the grade is the published checklist. How reviews work.

How the scores have moved

Verdicts

Doppler

Service tokens bound to one config, read-only by default, with --max-age expiry. Dynamic secrets and on-prem are Enterprise only, and Developer has no service accounts.

Pulumi ESC

An agent can start without a signup, because the Pulumi CLI creates a free ephemeral account that includes ESC, and the REST API has a public OpenAPI document. Audit logs, custom roles and approvals need the Pro edition at $400 a month, and no API rate limit with numbers was found.

Before you call either

Doppler

  1. Create a service token scoped to one config and read-only, then start the agent with doppler run --token $DOPPLER_TOKEN -- <cmd> so values never touch disk
  2. Start the MCP server with --read-only and --config as well as a scoped token; the server can't tell a token's permissions and would otherwise list write tools that fail
  3. Call /v3/configs/config/secrets/names when you only need names, and secrets/download?format=json for every value in one call
  4. On a 429 wait for the retry-after seconds; secret reads have their own limit, 120 a minute on Developer
  5. Run doppler configure flags disable analytics on build agents if you don't want CLI command usage reported

Pulumi ESC

  1. Use pulumi env, not esc. The standalone CLI stopped at v0.26.0 and gets no security fixes
  2. Read one value with pulumi env open <org>/<project>/<env> <property path> so the whole environment doesn't enter context
  3. Run tools with pulumi env run <env> -- <cmd>, which filters secret values from the command's output unless -i is set
  4. Set PULUMI_ACCESS_TOKEN for the SDKs. From 0.14.0 they no longer read the CLI login on disk
  5. Send Authorization: token <token> and Accept: application/vnd.pulumi+8 on REST calls, and expect 409 when an environment changed since it was read
  6. Relay the claim link an agent account prints. The account goes read-only after 72 hours and locks after 30 days. Don't run pulumi login first, which opens a browser

Questions

Which is better for AI agents, Doppler or Pulumi ESC?

Pulumi ESC scores 74.7 (BB) on agent readiness against Doppler's 71.9 (BB), and leads in 4 of 7 scored categories. Doppler leads on reliability and transparency & trust.

Do Doppler and Pulumi ESC need an API key?

Both take an API key or an OAuth sign-in.

Can an agent call Doppler and Pulumi ESC without installing anything?

Yes. Doppler has a hosted endpoint at https://api.doppler.com/v3 and Pulumi ESC at https://api.pulumi.com.

Other comparisons with Doppler or Pulumi ESC

Machine-readable

For companies

Do agents find, use and choose your tools?

An agent-readiness audit runs our probes, task suite and eight reviewer agents against your public and internal tools, and comes back with a scorecard, the transcripts of what failed, and a fix list in priority order. From $2,500, re-run included. We never take payment to move a rank. We do help companies earn one.