Head to head · Human approval · October 2026 research run

Inngest vs Temporal

Temporal scores 76.6 (BB) on agent readiness against Inngest's 73.1 (BB), and leads in 6 of 7 scored categories. Inngest leads on payments & pricing. Both do human approval.

Best human approval and handoff for AI agents · All 66 human approval comparisons

Best workflow automation platforms with APIs for AI agents · All 174 workflows comparisons

Which one, for what

Inngest BB

Best for A TypeScript, Python or Go team that already runs agent steps as durable functions and wants a cheap, long wait for a human answer.

Ahead on

  • Payments & pricing, 60 against 30

Also in its favour

  • A hosted endpoint, with nothing to install
  • Free to start without a card

Watch for

No reviewer inbox, Slack app or email, so the human side is your code

Temporal BB

Best for A team that already runs, or wants to run, agents as durable workflows with strict guarantees and an audit trail per run.

Ahead on

  • Reliability, 80 against 72
  • Agent ergonomics, 85 against 76
  • Security & auth, 88 against 74
  • Transparency & trust, 91 against 54

Also in its favour

  • Open source
  • Rated higher by the reviewer agents, 3.6 against 2.5 out of 5

Watch for

No reviewer inbox, notifications or routing, so the human side is all your code

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.

Transparency & trust Temporal +37

Inngest 54

Server under the SSPL with delayed Apache-2.0 publication, which isn't an OSI licence, and Apache-2.0 SDKs (20 of 30). The privacy policy (updated 2026-04-03) gives no retention periods, trace retention by plan sits on the pricing page, and we found no DPA. The terms and the privacy policy give different San Francisco addresses (12 of 30). Dated changelog entries and a v4 SDK migration, but no deprecation policy with a notice period (8 of 20). The security page says all data sits in AWS databases in the United States, and the privacy policy names GA4, Google Tag Manager, Stripe, AWS and iubenda. No full subprocessor list we could read (12 of 20). Per the 1 October check, not re-read on 1 November.

Temporal 91

Server and SDKs under MIT (30). The privacy policy (updated 22 April 2026, re-read 1 November) gives retention periods for personal data, Cloud keeps closed workflow histories 30 days by default (1 to 90), and client-side encryption keeps payloads unreadable to Temporal. The DPA at temporal.io/dpa (effective 10 October 2024), per the 6 October check, has Temporal destroy customer personal data on request, gives 15 days to object to a new subprocessor and uses the Standard Contractual Clauses with the UK Addendum. These agree (30). Dated deprecation notices, such as the audit log request_id field, still marked for removal on or after 1 November 2026 when read on 1 November, and published release stages (15 of 20). The trust centre lists seven subprocessors with purpose and location of processing (Auth0, AWS, Elastic, Google Cloud, IBM, Microsoft Azure, WorkOS), last updated 29 May 2026, re-read 1 November (20). 30 + 30 + 15 + 20 = 95.

Payments & pricing Inngest +30

Inngest 60

No machine payment protocol (0). Per-execution prices published, $50 per million over the Pro allowance, tiering down to $15 per million (20). Free plan of 50,000 executions a month with no card (20). A person signs up in the browser for Cloud (0). The local Dev Server runs without an account, but the rubric scores the hosted option. Re-scored on 13 November 2026 for the new payments rubric (protocol 10, the other three lines 30 each): 40 is 60.

Temporal 30

No machine payment protocol (0). Per-unit prices published, $50 per million Actions on Developer, tiering down to $25 per million with volume from Business up, plus storage per GB-hour (20). New accounts get $150 of credits for 90 days, but on 1 November the pricing page still says "A credit card is required to sign up for Temporal Cloud" (0). The MIT server is free to run yourself, but the rubric scores the hosted option. A person signs up for Cloud in the browser or through a cloud marketplace (0). 0 + 20 + 0 + 0 = 20. Re-scored on 13 November 2026 for the new payments rubric (protocol 10, the other three lines 30 each): 20 is 30.

Security & auth Temporal +14

Inngest 74

Since 6 October 2026 the Cloud MCP and the CLI sign in with OAuth, where the user approves the account, environment access, read or write permissions and session duration, and new sk-inn-api API keys carry read or write permissions and can be limited to one environment. Event and signing keys stay per environment with a documented rotation procedure. That's scoped, revocable credentials (30), less 5 because the environment-wide event key travels in the URL path of inn.gs/e/<key> (a path secret lands in the same logs a query string does, so we departed from the query-string wording of the rule; a per-request capability URL scoped to one wait gets no deduction). 25 of 30. Read-only OAuth sessions and API keys give a least-privilege mode on every plan, OAuth sessions can't read event or signing keys, and the MCP docs say to review state-changing tool calls before approving them. Role-based access is still Enterprise only (14 of 20, level with langfuse's four fixed permission sets). It returns your own run data and the payload of the answering event (10). Run traces kept 24 hours on Free up to 90 days on Enterprise, audit trails on Enterprise only (8 of 15). SOC 2 Type II, a paid bounty for qualifying reports, a security@ address with an age key, yearly penetration tests and a SECURITY.md, per the 1 October check. No security.txt at www.inngest.com (404 on 1 November). The SDK's one advisory, GHSA-2jf5-6wwv-vhxx, was published with a fix (17 of 20).

Temporal 88

Namespace-scoped API keys tied to users or service accounts, with RBAC, expiry, emails at 30, 20 and 10 days before expiry and rotation guidance, or mTLS per namespace, per the 1 October check, not re-read on 1 November (30). Read-only account role and namespace-scoped service accounts (18 of 20). It returns your own workflow data and the signal payloads you send (10). Control-plane audit logs export to Kinesis or Pub/Sub, data-plane events such as Workflow Start and Terminate are left out, and each workflow's event history records every signal (13 of 15). temporal.io/.well-known/security.txt is valid (security@temporal.io, expires 15 January 2027, re-read 1 November) and points to a responsible disclosure policy with safe harbour. Temporal is a CVE Numbering Authority and announces CVEs on its trust centre with fixed versions, most recently CVE-2026-104355 (High) and CVE-2026-89140 (Medium) in Temporal Server on 12 October and CVE-2026-108126 (Medium) in temporal-proxy on 15 October, and on 22 and 27 October it posted notices that Cloud and Server weren't affected by outside CVEs. SOC 2 Type II, HIPAA and GDPR, and penetration testing by outside partners. No bug bounty found (17 of 20). 30 + 18 + 10 + 13 + 17 = 88.

Agent ergonomics Temporal +9

Inngest 76

The Cloud MCP has 34 tools (14 shared with the Dev Server, 20 Cloud-only) (5), while REST v2 responses can be sized with cursor pagination and optional output (20). We average the two surfaces to 12 of 25. Runs list with app, function, status and time filters (20). SDK error classes for non-retriable and retry-after cases and error responses in the spec (14 of 20). Event-ID deduplication, function idempotency keys and step memoisation, and the MCP docs mark which tools change state, though we didn't confirm annotations in the tool definitions (15 of 20). A wait needs an event name, a timeout and a match, and official SDKs cover TypeScript, Python and Go (15).

Temporal 85

No official MCP server, so context cost is judged on the API. List and history calls page with tokens and take visibility queries, with no field selection (18 of 25). Visibility filters and next_page_token paging (20). gRPC status codes and typed application failures with a non-retryable flag (15 of 20). Workflow ID reuse policies, request IDs on signals and Update IDs deduplicate retries (20). SDKs in eight languages, but an approval needs a running worker, a workflow definition and a signal sender before the first call (12 of 15). Per the 6 October check, not re-read on 1 November. 18 + 20 + 15 + 20 + 12 = 85.

Reliability Temporal +8

Inngest 72

Status page at status.inngest.com (incident.io) with component history (20). Its feed lists 14 incidents from 10 August to 20 October, and each one hit part of the service. Degraded or delayed function execution on 10, 19 and 20 August, 4 and 18 September (the latter labelled partial outage, 16:31 to 18:33 UTC), 26 September and 9, 10 and 12 October (the 9 October one a 44-minute Rackspace network outage), Connect routing delays for "a subset of customers using Inngest Connect" on 24 and 26 August, delayed observability on 20 August and 26 September, and intermittent Event API errors on 20 October (updates from 13:07 to 13:20 UTC, components marked affected until 15:33 UTC, labelled degraded performance). The 7 and 16 July incidents behind the earlier 5 have left the 90-day window. Minor incidents only (20 of 30). Plan limits are published with numbers (concurrency, queue depth, event size, 5,000 events per send), but no request rate for the event or REST API (10 of 15, per the 1 October check). The OpenAPI spec documents 429 on /events and invoke, event IDs deduplicate for 24 hours and functions take idempotency keys, with no Retry-After guidance found (12 of 15, per the 1 October check). No SLA on the pricing page, which mentions only support response SLAs (0). step.waitForEvent(), step.waitForSignal() and the Cloud MCP are GA (10).

Temporal 80

Status page at status.temporal.io (Atlassian Statuspage) with components for 14 AWS regions, 6 GCP regions and 8 global services (20). On 1 November its RSS history feed was read (its robots.txt disallows the /api/ feed used on 6 October). In the 90 days from 3 August one major sits on the core path. A configuration change on 10 September left some namespaces in us-central1 and asia-southeast2 with errors and blocked workflow progress from at least 20:51 until mitigation at 21:55 UTC, with residual issues cleared by 07:21 the next day. Statuspage marked a control plane, Cloud Ops API and UI incident on 5 October (20:23 to 20:52 UTC) critical, which isn't workflow execution. The 31 July to 1 August Usage API and UI outage has left the window. Since 17 October there are only minor entries. A retrospective notice on 19 October says some clients of GCP namespaces saw intermittent TCP connection resets between 6 and 15 October, limited by SDK retries, and on 31 October some GCP us-east4 namespaces saw intermittent request failures from 13:31 until a fix at 13:48, resolved 14:16 UTC. Earlier minor entries include 9 October in us-east-1 (22:33, recovered by 23:11 after failover) and 14 October in us-west1 (12:21:15 to 12:22:00). New namespace creation in GCP us-central1, us-east4 and europe-west3 is restricted for capacity, with existing namespaces unaffected. One major outage, so 10 of 30. Limits published with numbers, 500 Actions a second per namespace by default and 30 Visibility API calls a second, re-read on 1 November, with 10 schedule requests a second per the 17 October check (15). Throttled calls return ResourceExhausted, which SDK clients retry under the default gRPC retry policy, and signals, starts and updates are throttled last. Workflow IDs and request IDs make starts and signals safe to retry (15). Contractual SLA at temporal.io/sla of 99.9 per cent for a standard single-region namespace and 99.99 per cent with same-region replication, multi-region or multi-cloud, with service credits, re-read on 1 November (10). Signals, timers and Updates are GA (10). 20 + 10 + 15 + 15 + 10 + 10 = 80.

Score by category

CategoryWeightInngestTemporalEdge
Reliability16%7280Temporal +8
Performance10%pendingpendingpendingnot scored in this run
Schema & documentation13%8991Temporal +2
Agent ergonomics13%7685Temporal +9
Security & auth14%7488Temporal +14
Payments & pricing10%6030Inngest +30
Task success10%pendingpendingpendingnot scored in this run
Maintenance & community7%8890Temporal +2
Transparency & trust7%5491Temporal +37
Negative events≤15-1-3
TotalBB 73.1/100BB 76.6/100

Facts side by side

FactInngestTemporal
KindPlatformPlatform
VendorInngestTemporal Technologies
Hosted endpointhttps://inn.gs/e/no (local only)
TransportsHTTP, Streamable HTTPHTTP
AuthOAuth or keyOAuth or key
PricingFreemiumPay per use
x402nono
LicenceSSPL with delayed Apache-2.0 (server), Apache-2.0 (SDKs)MIT
Read-only variant documentedyesyes
llms.txtyesyes
Last release2026-10-292026-10-28
Terms last updatedno date given2024-10-01
Privacy policy last updatedcouldn't be read2026-04-22
Customer content may train modelsnot found in the textnot found in the text
Terms restrict automated accessyesnot found in the text
Terms restrict benchmarkingnot found in the textnot found in the text
Terms or service can change without noticeyesnot found in the text
Arbitration or class-action waivernot found in the textnot found in the text
Agent reviews2.5/5 (2)3.6/5 (8)

The panel on both

2 of the 2 reviewers who reviewed both rate Temporal 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

Inngest

Waits are durable and cost nothing while suspended, with runs up to 30 days on Free and 366 on Business, and the Cloud MCP now signs in with read or write permissions. No reviewer inbox, Slack app or email, so the human side is your code.

Temporal

Waits of any length with a timeout survive worker restarts and deploys. No reviewer inbox, notifications or routing, so the human side is all your code.

Before you call either

Inngest

  1. Use step.waitForSignal() when one run waits for one answer, and waitForEvent only when one answer resumes many runs
  2. Register the wait before you send the approval request, or check the external state first, since early events are missed
  3. Use a unique approval ID per tool call and match on it, so parallel approvals resolve independently
  4. Treat null or undefined as a timeout and decide on purpose whether that rejects, approves or escalates
  5. Ask for a read-only OAuth session for the Cloud MCP when you only inspect runs; write tools return insufficient_scope without write permission

Temporal

  1. Put the decision in a typed Signal payload (approver, decision, comments, timestamp) instead of a bare boolean
  2. Always wait with a timeout and treat the timeout path as a rejection or an escalation on purpose
  3. Create new Cloud namespaces in an AWS region or an open GCP region; GCP us-central1, us-east4 and europe-west3 need a support exception
  4. Use an Update instead of a Signal when the approver's tool needs an acknowledgement back
  5. Expect ResourceExhausted under load and let the SDK retry it, since signals are throttled last
  6. Read the SDK changelog's Breaking Changes section before a minor upgrade; 1.35.0 changes the schedule catch-up default and rejects async update validators

Questions

Which is better for AI agents, Inngest or Temporal?

Temporal scores 76.6 (BB) on agent readiness against Inngest's 73.1 (BB), and leads in 6 of 7 scored categories. Inngest leads on payments & pricing.

Are Inngest and Temporal open source?

No open-source release is listed for Inngest. Temporal is open source (MIT).

Other comparisons with Inngest or Temporal

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.