Your AI coding agent can ask Rollbar how many users an error has reached, through get-metrics in our MCP server. Before it changes any code, the agent can see whether the error is growing day by day or only shows up on one browser or code version.
You can now revoke a Rollbar invite sent to the wrong address.
Rollbar Lines, our software factory, wrote the requirements, design, and code, ran acceptance checks, checked the release in production, validated UAT, and wrote the GTM, this post included.
https://t.co/sXAUxJJYBV
You can now send a teammate straight to the Improve card on your account dashboard with a link from its menu, the same way you already can with the Occurrences card. The Improve card's time-to-resolve table now follows the date range you pick for the card. https://t.co/m5iNUpXQHk
Ask your coding agent to "walk me through what the user did before the error," and it can answer from the clicks, inputs, and DOM changes Rollbar recorded in that session. The get-replay tool in our MCP server fetches the session replay for the agent. https://t.co/97btrHnoU1
@kaluops Not at the token level. Personal access tokens have read and write scopes, and write covers posting comments as well as PATCH /api/1/item/{id}, which changes status and assignee. The token can only do what its user can do in Rollbar, so the user's team access is the limit.
Your coding agent can now write what it found onto the Rollbar item, like "Investigated the regression; the fix is in PR #123." The post-item-comment tool in our MCP server posts the comment with your personal access token, under your name. https://t.co/dwKVQhPc30
Your AI agent can read the code that threw an error but not the request behind it. Our MCP server's get-item-details tool returns the error's latest occurrence with its request parameters and telemetry, so the agent sees the values that broke your code.
https://t.co/zTs351eaZD
@AiMarket24 Agreed, and it works in the other direction too. With a personal access token, the agent can leave a comment on the item saying what it tried, so whoever opens that error next, person or agent, sees it in the history before they reach for the same patch.
When a resolved error comes back, your AI agent can't see in the code what your team tried last time. The list-item-events tool in our MCP server reads that error's comments, status changes, and resolve notes, so the agent picks up where you left off.
https://t.co/MQzQfdqqWL
@Jauk_Air Agreed. If the error was resolved and came back, the agent sees in the status changes that the fix didn't hold. Past that, it's down to the comments. A note saying "workaround, remove after the DB upgrade" helps, and an agent with a personal access token can post one too.
We're not doing this step today, but it is on the roadmap. Currently, Resolve checks out the commit tied to the error, runs your tests and linters, and attaches the results to the PR. If the bug depends on data or config the sandbox lacks, a pass only proves what your tests cover.
Rollbar can now write the fix for a production error before anyone opens it. Agent Automation is on for every account. Set a rule, and Resolve tests a change for each matching error in an ephemeral environment and opens a pull request for your review.
https://t.co/Ml0UfomJJ1
An AI agent can read every file in your repo and still not know which lines are throwing errors in production. The list-items tool in our MCP server lets the agent search active Rollbar errors by text and environment before it edits anything.
https://t.co/7WIvJV4UFC
In June 2021, one valid customer config change made 85% of Fastly's network return errors. Fastly caught the outage in a minute but took 39 more to find the config. Send a config ID with every rollbar.js error to see which config the failures share.
https://t.co/imj4eeKjny
Sorry to hear that Greg, and thanks for emailing support. We'll pick it up from there and help you find a solution. Our MCP server’s tools don’t have access to create or delete an account, only the ability to read items and update comments. So installing it shouldn’t impact your account.
Your AI agent can resolve the Rollbar item once it fixes the bug. update-item in our MCP server records the version with the fix, and if your SDK sends code_version, Rollbar skips reactivation alerts until errors arrive from that version or later.
https://t.co/i2lgQgr4qa
Your AI agent spent an afternoon patching a bug Friday's deploy already fixed. The code gave no sign the errors stopped. Our get-version MCP tool looks up a git SHA and shows when its errors first and last appeared, so the agent can check before patching.
https://t.co/eSsUDdzQya
Your AI agent can read the retry loop in your code, but it can't see how often that loop gives up in production. Our MCP server's list-occurrences tool gives the agent a timestamped record of each failure, so it can count them before editing the backoff.
https://t.co/1ruCGjOAfZ
@botgatheringcom Agreed. Anyone holding a read-only token can still see everything in the payload, and error payloads pick up auth headers and env vars more often than anyone likes. So scrub secrets in the app before the error is sent, and someone with a leaked read token only sees [REDACTED].
LiteLLM's MCP endpoint accepted forged tokens, a flaw CISA lists as exploited (CVE-2026-59822). Upgrade to 1.84.0 or later. If you run our MCP server behind LiteLLM, a read-scope token means a forged session can read your errors but can't change them.
https://t.co/DE94YkP1WI