Skip to main content
kcapProduct page

kcap Signals

What engineering leaders and their teams report about kcap, with every finding tied to its evidence.

Read the findings

What the evidence says

Whose sources count

Practitioner sentiment, all practitioners

Unscored until 20 sources from 15 authors

How this figure is made

Unscored until 20 sources from 15 authors

0 sources from 0 authors so far. GILD prints practitioner sentiment at 20 from 15. Run a pilot on one production app to measure which failures it surfaces and how long tracing them takes.

Analyzed
16 total sources
Top praise
For all practitioners, no praise theme reaches 5 sources from 3 authors.
Top concern
For all practitioners, no concern theme reaches 5 sources from 3 authors.

Jun 2026 to Nov 2026.

11 sources from practitioners, by month and sentiment.

The themes below keep their own minimum of 5 sources from 3 authors. GILD found no independent review or deployment account of kcap as of Nov 4, 2026. Kurrent's staff and channels wrote 10 of the 11 records listed, a LinkedIn post came after Kurrent reached out to its author, and outside bug reports were written with coding agents, so none counts. The findings below rest on Kurrent's own statements.

Positive, mixed and negative sources

See the evidence

Findings

Counted for All practitioners, from 11 sources. Change group

What practitioners praise

Every capability practitioners describe positively, ranked by how many sources say so. Open one to read who said it.

For all practitioners, no capability reaches 5 sources from 3 authors.

    What concerns them

    Every concern practitioners raise, most raised first.

    For all practitioners, no concern reaches 5 sources from 3 authors.

      Where the evidence agrees

      Statements this group's sources agree on, then the ones close to agreement.

      For all practitioners, no statement reaches 5 sources from 3 authors.

      Where it disagrees

      Questions the sources split on, with the count on each side.

      For all practitioners, no question reaches 5 sources from 3 authors, with 2 on each side.

        The evidence

        Whose experience this is

        The three groups, then every practitioner by the context they work in, and how emphasis changes with scale.

        Three groups drawn from the same sources. Choose a group to select whose experience counts. Within that group, each author counts once in the practitioner sentiment, with their latest view. Signals open on All practitioners. Leaders and teams have 0 accounts; GILD prints their sentiment from 20.

        • Engineering leaders and teams

          10 sources. This group counts the accounts of engineering teams and the leaders setting adoption across them.

        • Individual builders

          1 sources. This group counts solo developers and independent builders working on their own.

        • All practitioners

          11 sources. This gives the broadest market read, but combines work settings that can produce materially different conclusions.

        11 sources from practitioners, by the context they work in. Filled rows count for this group.

        1. Individual practitionerSolo developers and independent builders1 sources, counted here

          Emphasizes One vendor-solicited LinkedIn post, listed with its tie and counted toward nothing

        2. Very small team2 to 10 engineersNo sources from this setting yet
        3. Professional team11 to 50 engineers, or a named company's team10 sources, counted here

          Emphasizes 10 records by Kurrent's own staff and channels, listed with their tie and counted toward nothing. GILD found no independent account from a named company's team as of Nov 4, 2026

        4. Technical leadershipLeaders setting adoption across teamsNo sources from this setting yet
        5. Organization-wideEnterprise or company-wide rolloutNo sources from this setting yet

        What changes with scale. GILD found no independent account of kcap at any team size as of Nov 4, 2026, so no setting has a read yet. Kurrent's pages set the terms a pilot tests: seats, invites, exclusions and who pays for headless calls.

        Where the sources come from

        Where this group's sources were published.

        Counted for All practitioners, from 11 sources. Change group

        16 sources for this group, 5 of them documentation

        • Developer Communities8, 50% of 16 sources
        • kcap Documentation5, 31% of 16 sources
        • Engineering Blogs & Reviews2, 13% of 16 sources
        • Videos & Podcasts1, 6% of 16 sources

        Kurrent renamed Kurrent Capacitor to kcap in October 2026 and says the product is the same, so a record about Kurrent Capacitor describes kcap.

        See sources by origin

        The record

        Who each source is, how it knows, whether it still applies and what it counts toward.

        Newest first, on its own page. Every count above opens its sources there.

        Read all 16 sources

        About this read

        What changed, and how we count

        What changed

        • Nov 4, 2026: first edition, from a search of Hacker News, GitHub, X, LinkedIn, YouTube, conference schedules, podcasts and the press, and a partial search of Reddit, with every page stored and checked against its stored copy by a second reviewer.
        • The one LinkedIn post found was written after Kurrent's team reached out to its author, so it is listed with that tie and counts toward nothing.
        • Outside users' bug reports on kurrent-io/kcap were written with coding agents, so none of them counts. Kurrent's own replies on two of them are fact checks under Service Incidents.
        • Fact checks quote Kurrent's pricing page, llms.txt, license and kcap documentation as stored on the review date.
        • Nov 4, 2026: added 10 records by Kurrent staff and Kurrent's own channels, from Hacker News, X, Medium, LinkedIn, YouTube, Kurrent's blog and kurrent-io/kcap issues, at most two per author, each listed with its tie and counted toward nothing. Staff statements of defects on eight kurrent-io/kcap issues are fact checks under Service Incidents, in Kurrent staff's words.
        • Nov 4, 2026: Kurrent's AI providers page joined the documentation records, since its default judge models include Claude Sonnet 4.6.

        Method

        A running review of what engineering teams report about kcap, from individual builders to professional teams. Each finding links to its sources, and each conflict shows both sides.

        Counts cover only the sources GILD reviewed.

        Who counts

        Engineering leaders and teams
        Very small team, Professional team, Technical leadership, Organization-wide
        Individual builders
        Individual practitioner
        All practitioners
        Everyone in the leaders and teams and individual builders groups

        Choose a group to select whose experience counts. Within that group, each author counts once in the practitioner sentiment, with their latest view. GILD files each source in a group from how the account describes its author. A role or employer marked self-described is what the source states about themselves. Documentation checks facts for every group and never counts toward sentiment.

        How the numbers work

        • Overall signal. Positive votes minus negative ones, per 100 votes from the sources that take a side. Each author counts once, with their latest view, whoever they are and wherever they posted. The older weighted reading, where accounts GILD matched to a named person count three times when GILD lists the author among engineering leaders and twice otherwise, is shown beside it as a second view. Where an author wrote more than one source, the line beside the split gives the votes, one per author, so the score follows from it. When one author's sources from the same day disagree, that author counts as mixed, and a half point rounds away from zero. Sources with no opinion are listed and not counted. Strong positive is +50 or higher, positive +20 to +49, mixed −19 to +19, negative −20 to −49, strong negative −50 or lower.
        • Unscored. A group gets a score only with 20 sources from 15 authors. Below that, one source moves the score about 5 points, so the read says "Unscored until 20 sources from 15 authors" and shows its counts.
        • Independence. A source whose author works for the vendor, for one of its partners or programs, or for a competitor, and a page the vendor published or ran, is listed with that tie and counts toward nothing. One author counts on at most two records per product. The line under the score says how many authors and how many separate discussions the counted sources come from: comments in one thread are one discussion.
        • Praise and concerns. How many of the sources that discuss a theme praise it or raise it.
        • Consensus. A statement joins when 20% of the group's sources support it and fewer contradict it. Contradicting sources are counted separately. Below that line a statement shows as building. At 40% or more it reads Strong agreement; below, Recurring point.
        • Disagreements. The count on each side. GILD names no winner and gives no share of all users.
        • Minimum evidence. A finding needs 5 sources from 3 authors; a disagreement also needs 2 on each side. Smaller counts stay in the table below, marked under the minimum.

        Labels on a source

        A source's date and whether it still applies are separate: a recent post can describe an old version, and an old one can still hold.

        Documented
        Stated in official documentation.
        Reported
        A practitioner's account of their own use.
        Measured
        Backed by the source's own measurements, such as a comparative test or tracked team metrics.
        GILD analysis
        GILD's check of a claim against official documentation.
        Current
        Describes the product as it is now.
        Still applicable
        Older, and the behavior or lesson still holds.
        Needs a recheck
        May depend on a plan, version or setting that changed since it was written. It keeps counting until GILD rechecks it, since the rest of the account usually holds, and its card says so.
        Out of date
        No longer accurate. Kept in the record; counts toward nothing current.

        Flagged for a recheck, of the sources that take a side: leaders and teams 0 of 0, individual builders 0 of 0, all practitioners 0 of 0.

        Limits

        • Kurrent is a GILD event partner, and any record tied to Kurrent is listed with its tie and counts toward nothing. GILD searched Hacker News, GitHub, X, LinkedIn, YouTube, conference schedules, podcasts and the press on Nov 4, 2026, and found no independent review or deployment account of kcap. Reddit's own search could not be reached that day, and an archive of Reddit answered 11 of 48 searches, all with no result. Kurrent's staff and channels supply 10 of the 11 records listed, at most two per author. Kurrent staff's own statements of defects in kcap, on eight kurrent-io/kcap issues, are fact checks under Service Incidents. Outside users filed bug reports on kurrent-io/kcap, written with coding agents, so GILD counts none of them. Fact checks quote Kurrent's own pages and its staff's words on kcap issues.
        • The records run from Kurrent Capacitor, the name kcap carried until Kurrent renamed it in October 2026, to kcap v1.4.1-beta.3 in Kurrent's own issues. Kurrent's docs call kcap 1.3.1 current stable, and GitHub lists v1.4.0 as its newest release that is not a prerelease, published Oct 27, 2026.

        Every count, and the sources behind it

        8 findings for 3 groups

        Each row opens its sources for the group you are reading.

        FindingEngineering leaders and teamsIndividual buildersAll practitionersSources
        Current readCurrent sentiment0 positive, 0 mixed, 0 negative0 positive, 0 mixed, 0 negative0 positive, 0 mixed, 0 negativeSources
        PraiseShared Session Memory0 of 0 positive (under the minimum)0 of 0 positive (under the minimum)0 of 0 positive (under the minimum)Sources
        PraiseIntegrations & Model Providers0 of 0 positive (under the minimum)0 of 0 positive (under the minimum)0 of 0 positive (under the minimum)Sources
        PraiseEvaluations & Usage Analytics0 of 0 positive (under the minimum)0 of 0 positive (under the minimum)0 of 0 positive (under the minimum)Sources
        PraisePrivacy & Access Controls0 of 0 positive (under the minimum)0 of 0 positive (under the minimum)0 of 0 positive (under the minimum)Sources
        ConcernSeats, Editions & Platform Limits0 of 0 raise it (under the minimum)0 of 0 raise it (under the minimum)0 of 0 raise it (under the minimum)Sources
        ConcernService Incidents0 of 0 raise it (under the minimum)0 of 0 raise it (under the minimum)0 of 0 raise it (under the minimum)Sources
        Current readEvidence mix15 sources6 sources16 sourcesSources

        Fact checks

        GILD's checks of claims against official documentation, kept apart from the source counts.

        • Shared Session Memory. A small hook in each coding agent sends the session to the kcap workspace as it happens: the prompts, the agent's replies, every tool call and its result, and the commits it made. kcap docs: What kcap is, checked Nov 4, 2026
        • Shared Session Memory. kcap's docs list Claude Code, Codex, Cursor, Gemini CLI, Copilot CLI, Kiro, OpenCode, Pi, Antigravity, Muse Code, Grok Build and Mistral Vibe. Every one is stored and importable, and most can also run as hosted agents. kcap docs: What kcap is, checked Nov 4, 2026
        • Shared Session Memory. All kcap MCP servers run locally as kcap mcp <name> and turn the agent's tool calls into authenticated requests against the workspace. Reads are filtered by the user's session visibility, and writes act as the user. kcap docs: MCP tools and skills, checked Nov 4, 2026
        • Shared Session Memory. kcap injects a session-start briefing into the agent. By the server defaults, which an operator can change, an approved document may be up to 4,000 characters, at most three documents and eight facts are included, and the whole briefing is capped at 16,000 characters. kcap docs: Session-start context, checked Nov 4, 2026
        • Shared Session Memory. A hosted agent runs on a machine the customer connects, such as a laptop, a workstation or a shared box, through the kcap daemon. Launched from the browser, it runs in a fresh git worktree, and the daemon's Agent capacity setting caps how many run at once, with 0 meaning no limit. kcap docs: Hosted agents, checked Nov 4, 2026
        • Shared Session Memory. A Page published from kcap can be up to 4 MB and keep up to 50 versions. kcap docs: Pages, checked Nov 4, 2026
        • Integrations & Model Providers. kcap's GitHub App and Linear app are each installed once for the workspace, with no personal token involved. kcap docs: Integrations, checked Nov 4, 2026
        • Integrations & Model Providers. One Slack app shows kcap session and agent links as cards in Slack and, where notifications are on, delivers direct messages. kcap docs: Integrations, checked Nov 4, 2026
        • Integrations & Model Providers. Titles, outcome summaries, embeddings and server evaluations run on Kurrent AI on every plan, and each AI feature can also run on a provider the customer brings: Anthropic, OpenAI, OpenRouter or Cohere. The Insights agent, work items and projects need Team or Enterprise, and a bring-your-own key does not turn them on. The local edition has no Kurrent AI. kcap docs: AI providers and models, checked Nov 4, 2026
        • Integrations & Model Providers. Embeddings for memories, fact clustering and work-item matching can use Voyage (default voyage-4-lite), OpenAI text-embedding-3-small, Kurrent AI, or on the local edition a built-in model or Ollama mxbai-embed-large. kcap docs: AI providers and models, checked Nov 4, 2026
        • Integrations & Model Providers. On the local edition, a key from Anthropic, OpenAI, Cohere or OpenRouter turns on titles, outcomes, summaries and server-side evaluations, and the docs say the calls go straight from the Mac to that provider and nothing passes through Kurrent. The other option runs that work through Claude Code or Codex on the Mac, with sonnet as its default model. kcap docs: Local edition, checked Nov 4, 2026
        • Evaluations & Usage Analytics. The kcap evaluation judge model is set per vendor, with defaults claude-sonnet-4-6, gpt-5-mini, command-a-03-2025 and google/gemini-2.5-flash. kcap docs: AI providers and models, checked Nov 4, 2026
        • Evaluations & Usage Analytics. kcap evaluation runs default to a 30-minute run timeout and a 5-minute judge timeout per question, reserve 2 minutes for the retrospective, and substitute a trace of up to 400,000 characters into a prompt. kcap docs: AI providers and models, checked Nov 4, 2026
        • Evaluations & Usage Analytics. kcap computes cost when it is viewed, from the tokens the agent reported and the model's price synced from models.dev. A model with no known price shows no cost, which the page labels unknown. kcap docs: Tokens and cost, checked Nov 4, 2026
        • Evaluations & Usage Analytics. On Team and Enterprise plans and on the local edition, the coding agent can query the governed views Insights uses: counts, tokens, cost, tool use and evaluation scores, never transcript text. kcap docs: What your agent can do, checked Nov 4, 2026
        • Privacy & Access Controls. An excluded repository or folder is never stored by kcap, live or on import, whatever else is configured. The one exception on import is Antigravity, whose history records no working directory, while a configured allow list skips it. kcap docs: Keep private repos out, checked Nov 4, 2026
        • Privacy & Access Controls. Members' sessions default to org_public, which the setup docs define as readable by the organisation for the organisation's own repositories. An admin sets the org-visible list of GitHub organisation or user names, and a blank list uses the connected GitHub account. The page adds "Without it, org_public sessions stay private." kcap docs: Set up the workspace, checked Nov 4, 2026
        • Privacy & Access Controls. Deleting a kcap session removes its data from the server, including its subagents, and cannot be undone. The agent's own transcript file on the machine stays, and anonymous counts such as those in Insights are recomputed. kcap docs: Delete a session, checked Nov 4, 2026
        • Privacy & Access Controls. kcap sessions are retained until deleted: a member can delete their own sessions and an admin can delete any session. Kurrent says workspace data is never used to train models, and never sold or shared. kcap docs: Data security and privacy, checked Nov 4, 2026
        • Privacy & Access Controls. kcap's pricing page lists SSO and directory integration for Enterprise, and not for Local or Team. kcap pricing, checked Nov 4, 2026
        • Privacy & Access Controls. Members sign in to the kcap web app with the provider the workspace uses: single sign-on with a work email, or GitHub for workspaces that use the GitHub App. kcap docs: Your account and preferences, checked Nov 4, 2026
        • Privacy & Access Controls. kcap machine credentials let CI runners, build boxes and agent sandboxes store sessions without a person signing in. They are available on workspaces that sign in through single sign-on, need the organisation owner or admin role and kcap administrator rights to create, and are never an administrator or a project member. kcap docs: Machine credentials, checked Nov 4, 2026
        • Privacy & Access Controls. KCAP_TELEMETRY=0 or DO_NOT_TRACK=1 turns off the kcap CLI's usage data, and the desktop switch then defers to it. kcap docs: Configuration, checked Nov 4, 2026
        • Seats, Editions & Platform Limits. kcap Team has 100 seats and Enterprise unlimited seats. Seats are active members plus pending invitations, and at the limit invitations are refused until a seat is freed. kcap docs: Plan and seats, checked Nov 4, 2026
        • Seats, Editions & Platform Limits. The kcap local edition has no teams, projects, sharing or machine credentials, and no sign-in. The Keep private repos out docs say the local server is for a machine with one user, and anyone else signed in to the same machine can open its web app and act as its admin. kcap docs: Local edition, checked Nov 4, 2026
        • Seats, Editions & Platform Limits. The local kcap agent commands are not available on Windows, and the docs say to launch from the browser or the desktop app. kcap docs: Hosted agents, checked Nov 4, 2026
        • Seats, Editions & Platform Limits. On Windows, a kcap app update stops the daemon and interrupts running agents mid-turn, and the Windows daemon keeps the old version until it is restarted. On the stable channel, CLI updates are capped at the version the workspace server runs. kcap docs: Keep kcap up to date, checked Nov 4, 2026
        • Seats, Editions & Platform Limits. Kurrent's terms say free workspaces that show no usage for 30 days or more may be deprovisioned after notice, with captured content available for export for 30 days after. Kurrent terms, checked Nov 4, 2026
        • Seats, Editions & Platform Limits. kcap replaces records over 4 MiB with a marker, and declared plan documents are limited to 256 KB each. kcap docs: Data security and privacy, checked Nov 4, 2026
        • Seats, Editions & Platform Limits. Kurrent's llms.txt prints a Team tier for up to 100 seats at US$15 per person per month billed annually (US$18 monthly), while the pricing page's structured data lists Team at US$30 a month for one seat, or US$100 a month for five seats plus US$25 per additional seat. Kurrent llms.txt, checked Nov 4, 2026
        • Seats, Editions & Platform Limits. On the kcap pricing page the Local and Team buttons read Request an invite and the Enterprise button reads Talk to us, while Kurrent's llms.txt describes a self-serve free tier from kcap setup or the browser. kcap pricing, checked Nov 4, 2026
        • Seats, Editions & Platform Limits. Kurrent's Redeem page marks Insights Not included for the free local install, the pricing page marks the Insights agent Your model API key for Local, and the local edition docs say the Insights agent is absent there. kcap: Redeem (free local install), checked Nov 4, 2026
        • Seats, Editions & Platform Limits. The kcap desktop app docs say the app runs on Apple silicon Macs today, with Windows and Linux to follow, while Kurrent's GitHub README for the client lists Windows 10 or later on x64. kcap docs: Desktop app, checked Nov 4, 2026
        • Service Incidents. On Aug 26, 2026 realtonyyoung of Kurrent wrote on kcap issue #642 that the "Root cause was on our side": during an infrastructure-degraded window from Aug 9 to 22, "session-end signals stopped being applied to the tenant's query index, so ended sessions vanished from the list while remaining searchable and openable". The comment adds that "no session data was ever lost" and that "The Aug 22 server upgrade rebuilt the index". kurrent-io/kcap issue #642: reply by realtonyyoung, Kurrent Principal Solutions Architect (Aug 26, 2026), checked Nov 4, 2026
        • Service Incidents. On Jun 11, 2026 alexeyzimarev of Kurrent replied "Thank you! Fixing now." on kcap issue #140, a report that kcap 0.6.14 and 0.6.15 broke Claude Code subagent calls on Windows, and wrote on Jun 13, 2026 "Released in 0.6.16". kurrent-io/kcap issue #140, the report and replies by alexeyzimarev, Kurrent Head of Engineering (Jun 11 and Jun 13, 2026), checked Nov 4, 2026
        • Service Incidents. Tony Young of Kurrent filed kcap issue #945 on Sep 14, 2026 for a customer whose Windows daemon often stopped and needed a manual restart, with repositories gone after some restarts. On Nov 2, 2026 Lokhesh Ujhoodha of Kurrent replied that both symptoms have had fixes, PRs #1279 and #1277, which kcap v1.3.0 lists, and asked for a check on 1.4.x. kurrent-io/kcap issue #945, by Tony Young of Kurrent, and Lokhesh Ujhoodha's reply (Sep 14 and Nov 2, 2026), checked Nov 4, 2026
        • Service Incidents. Tony Young of Kurrent wrote on kcap issue #1275, Oct 1, 2026, that after a failed read of repos.json the next add replaced the whole saved list with one entry, reproduced in a unit test on macOS. kcap v1.3.0 lists the fix, PR #1277. kurrent-io/kcap issue #1275, by Tony Young of Kurrent (Oct 1, 2026), checked Nov 4, 2026
        • Service Incidents. Tony Young of Kurrent wrote on kcap issue #1274, Oct 1, 2026, that the documented update step stopped a Windows service daemon until the user logged off and on again, reasoned from the code and not yet observed on a Windows machine. kcap v1.3.0 lists the fix, PR #1279. kurrent-io/kcap issue #1274, by Tony Young of Kurrent (Oct 1, 2026), checked Nov 4, 2026
        • Service Incidents. Alexey Zimarev of Kurrent wrote on kcap issue #1155, Sep 25, 2026, that at about 10x slowdown a legitimate large tool result is lost to the 100 ms redaction deadline, a loss computed with no observed count. On Oct 26, 2026 Norton Andreev of Kurrent wrote that the data loss is fixed on main. Kurrent's v1.3.0 release lists the fixes, PRs #1366 and #1404. kurrent-io/kcap issue #1155, by Alexey Zimarev of Kurrent, and Norton Andreev's status note (Sep 25 and Oct 26, 2026), checked Nov 4, 2026
        • Service Incidents. Tony Young of Kurrent wrote on kcap issue #859, Sep 10, 2026, that a service-installed daemon on kcap 1.0.0 for macOS came up 13 minutes after spawn while status said running. kcap v1.0.2 lists the pipe fix, PR #869. kurrent-io/kcap issue #859, by Tony Young of Kurrent (Sep 10, 2026), checked Nov 4, 2026
        • Service Incidents. Tony Young of Kurrent wrote on kcap issue #714, Aug 31, 2026, that on kcap 0.11.35 a daemon restart left a hosted Cursor session as a zombie that dropped the user's input. The issue closed Sep 1, 2026, and no stored page names a fix version. kurrent-io/kcap issue #714, by Tony Young of Kurrent (Aug 31, 2026), checked Nov 4, 2026
        • Service Incidents. Lokhesh Ujhoodha of Kurrent wrote on kcap issue #2226, Nov 2, 2026, that kcap's SessionEnd hook is cancelled on a Windows machine without Git Bash, so claude -p still prints a hook failure there. The pre-release kcap v1.4.1-beta.3 lists the partial fix, PR #2267. kurrent-io/kcap issue #2226, Lokhesh Ujhoodha of Kurrent (Nov 2, 2026), checked Nov 4, 2026
        • Service Incidents. Norton Andreev of Kurrent wrote on kcap issue #1851, Oct 22, 2026, that after a daemon restart the server could record an unrelated terminal session as a hosted Claude agent's continuation. The issue closed Oct 26, 2026, and no stored page names the fix. kurrent-io/kcap issue #1851, by Norton Andreev of Kurrent (Oct 22, 2026), checked Nov 4, 2026