# CoreSpeed Blog

Source: https://corespeed.io/blog

Notes from building the harness AI agents run on — connectors, memory, budgets, billing, and the engineering in between.

## Posts

- [Xbrain: Give Your Agent a Sense of Your Taste](https://corespeed.io/blog/introducing-xbrain) — 2026-10-08, by Milton Yan. Turn your X bookmarks into searchable design context for your agent. (Markdown: https://corespeed.io/blog/introducing-xbrain.md)
- [A meeting brief from your mail, calendar and notes, in one request](https://corespeed.io/blog/a-meeting-brief-from-mail-calendar-and-notes) — 2026-10-07, by CoreSpeed. How an agent builds a meeting brief from Gmail, Google Calendar and Notion through private connections, remembers your format, and never holds a token. (Markdown: https://corespeed.io/blog/a-meeting-brief-from-mail-calendar-and-notes.md)
- [A support agent that issues refunds, with a person on the big ones](https://corespeed.io/blog/support-refunds-with-an-approval-step) — 2026-10-07, by CoreSpeed. A worked example. A support agent reads the ticket, checks the charge in Stripe and issues the refund. Refunds over $100 pause for a person in Slack. (Markdown: https://corespeed.io/blog/support-refunds-with-an-approval-step.md)
- [Built-in web search vs wiring a search API into your agent](https://corespeed.io/blog/built-in-web-search-vs-connecting-a-search-api) — 2026-10-07, by CoreSpeed. CoreSpeed's web tools return ranked sources and clean page text with no provider account, billed per request. A direct provider key gives more knobs. (Markdown: https://corespeed.io/blog/built-in-web-search-vs-connecting-a-search-api.md)
- [Competitive research: search for sources, then read what matters](https://corespeed.io/blog/competitive-research-with-search-and-scrape) — 2026-10-07, by CoreSpeed. A playbook for an agent that finds competitor sources with web__search, reads the few pages that matter with web__scrape, and writes the brief to Notion. (Markdown: https://corespeed.io/blog/competitive-research-with-search-and-scrape.md)
- [Connect Claude Code to Notion, Slack and GitHub with one MCP server](https://corespeed.io/blog/connect-claude-code-to-your-apps-with-one-mcp-server) — 2026-10-07, by CoreSpeed. Add the CoreSpeed MCP server to Claude Code at user scope, sign in once, connect Notion, Slack and GitHub in the dashboard, and verify with manage__whoami. (Markdown: https://corespeed.io/blog/connect-claude-code-to-your-apps-with-one-mcp-server.md)
- [Continue a task in a different agent without re-explaining](https://corespeed.io/blog/continue-a-task-in-a-different-agent) — 2026-10-07, by CoreSpeed. What carries over when you move a task from Claude Code to Codex through CoreSpeed, memory, connections and policy, and what does not, the conversation itself. (Markdown: https://corespeed.io/blog/continue-a-task-in-a-different-agent.md)
- [CoreSpeed FAQ: plans, credits, clients, security, what it does not do](https://corespeed.io/blog/corespeed-faq) — 2026-10-07, by CoreSpeed. Short answers on what CoreSpeed is, which agents work, what is free, where tokens live, who approves, whether it runs agents, and how to leave. (Markdown: https://corespeed.io/blog/corespeed-faq.md)
- [Exactly once: how an approved agent action runs with frozen arguments](https://corespeed.io/blog/exactly-once-approvals) — 2026-10-07, by CoreSpeed. A held call is frozen, the agent waits on one id, and approval hands the call to that waiting request once. Re-sending the call makes a second card. (Markdown: https://corespeed.io/blog/exactly-once-approvals.md)
- [Generating launch images and short videos from an agent](https://corespeed.io/blog/generate-launch-assets-with-media-tools) — 2026-10-07, by CoreSpeed. How an agent picks a hosted model, generates a launch image, chains it into a short video with an art_ id, and collects signed URLs. Billed only after success. (Markdown: https://corespeed.io/blog/generate-launch-assets-with-media-tools.md)
- [How memory search works: hybrid retrieval, excerpts and result budgets](https://corespeed.io/blog/how-memory-search-works) — 2026-10-07, by CoreSpeed. memory__search_memory fuses exact and relaxed lexical matching, entity aliases and vector similarity into one ranking, then bounds each row and the result. (Markdown: https://corespeed.io/blog/how-memory-search-works.md)
- [How to add one MCP server to Cursor for apps, memory and web search](https://corespeed.io/blog/add-corespeed-to-cursor) — 2026-10-07, by CoreSpeed. Add CoreSpeed to ~/.cursor/mcp.json, click Needs login, and Cursor gets your connected apps, durable memory, live web search and media tools from one server. (Markdown: https://corespeed.io/blog/add-corespeed-to-cursor.md)
- [How to connect ChatGPT to your apps with a custom connector](https://corespeed.io/blog/use-corespeed-in-chatgpt-developer-mode) — 2026-10-07, by CoreSpeed. Turn on Developer mode in ChatGPT (Plus and Pro on the web), create a connector for api.corespeed.io/mcp with OAuth, and sign in to reach your apps and memory. (Markdown: https://corespeed.io/blog/use-corespeed-in-chatgpt-developer-mode.md)
- [How to connect Codex to your apps and give it memory](https://corespeed.io/blog/connect-codex-to-your-apps-and-memory) — 2026-10-07, by CoreSpeed. Register CoreSpeed in Codex with codex mcp add, sign in, connect your apps, and add memory rules to AGENTS.md so Codex recalls decisions across sessions. (Markdown: https://corespeed.io/blog/connect-codex-to-your-apps-and-memory.md)
- [How to connect two accounts of the same app to one agent](https://corespeed.io/blog/connect-two-accounts-of-the-same-app) — 2026-10-07, by CoreSpeed. Connect two Slack workspaces or two X handles under one connector, give each an alias, and pass the account argument so each call lands in the right one. (Markdown: https://corespeed.io/blog/connect-two-accounts-of-the-same-app.md)
- [How to give your agent memory that survives sessions and clients](https://corespeed.io/blog/give-your-agent-memory-across-sessions) — 2026-10-07, by CoreSpeed. Two standing rules, a scope choice, and a habit of saving decisions give any agent memory that outlives the session and the client it ran in. (Markdown: https://corespeed.io/blog/give-your-agent-memory-across-sessions.md)
- [How to register your own MCP server for your whole organization](https://corespeed.io/blog/register-your-own-mcp-server-as-an-org-connector) — 2026-10-07, by CoreSpeed. An org admin registers any HTTPS MCP server as a remote connector; its tools join tools/list under the org__ prefix with the same sign-in, caps and trail. (Markdown: https://corespeed.io/blog/register-your-own-mcp-server-as-an-org-connector.md)
- [How to run a headless agent or CI job with a CoreSpeed API key](https://corespeed.io/blog/api-keys-for-headless-agents-and-ci) — 2026-10-07, by CoreSpeed. Choose a member key (sk-cs-) or an agent key (sk-csa-), read it from CORESPEED_API_KEY, cap its monthly spend, and verify the key with memory__list_memory. (Markdown: https://corespeed.io/blog/api-keys-for-headless-agents-and-ci.md)
- [How to set a monthly budget for an agent](https://corespeed.io/blog/set-a-monthly-budget-for-an-agent) — 2026-10-07, by CoreSpeed. Give the agent its own key with a monthly spend cap in credits, know what the refusal looks like, and read what each run cost in Activity. (Markdown: https://corespeed.io/blog/set-a-monthly-budget-for-an-agent.md)
- [How to share one app connection with your whole team](https://corespeed.io/blog/share-an-app-connection-with-your-team) — 2026-10-07, by CoreSpeed. Connect an app once as a shared account so every teammate's agent can act through it. Who can share, who can remove, and when to keep a connection private. (Markdown: https://corespeed.io/blog/share-an-app-connection-with-your-team.md)
- [How to use your apps and memory inside claude.ai and Claude Desktop](https://corespeed.io/blog/use-corespeed-in-claude-ai-and-claude-desktop) — 2026-10-07, by CoreSpeed. Add CoreSpeed as a custom connector on the Pro, Max, Team or Enterprise plan. It follows your Claude account to desktop and mobile with your apps and memory. (Markdown: https://corespeed.io/blog/use-corespeed-in-claude-ai-and-claude-desktop.md)
- [How to write your first approval policy (with examples)](https://corespeed.io/blog/write-your-first-approval-policy) — 2026-10-07, by CoreSpeed. Write three kinds of sentence (fine, ask, never), pick a default behavior, run Test policy, connect Slack and turn the gate on. Reads are never gated. (Markdown: https://corespeed.io/blog/write-your-first-approval-policy.md)
- [HTTP 200 with isError true: the most common MCP integration bug](https://corespeed.io/blog/http-200-with-iserror-true) — 2026-10-07, by CoreSpeed. A CoreSpeed tool call can fail inside an HTTP 200. How the two error layers differ, which codes to retry, and which ones need a person instead. (Markdown: https://corespeed.io/blog/http-200-with-iserror-true.md)
- [Linking memories: a small graph that keeps decisions connected](https://corespeed.io/blog/a-memory-graph-with-links) — 2026-10-07, by CoreSpeed. Memory links are directed edges with a kind and optional weight. They are idempotent to write, visible only when both ends are, and capped at 1,000 per memory. (Markdown: https://corespeed.io/blog/a-memory-graph-with-links.md)
- [Market scans with listing tools: Zillow, Airbnb, TripAdvisor and Maps](https://corespeed.io/blog/real-estate-and-travel-scans-with-listing-tools) — 2026-10-07, by CoreSpeed. Read Zillow, Airbnb, TripAdvisor and Google Maps as structured records. Each call collects live in 10 to 60 seconds and bills only the rows it gathered. (Markdown: https://corespeed.io/blog/real-estate-and-travel-scans-with-listing-tools.md)
- [Member key vs agent key vs browser sign-in: which to use](https://corespeed.io/blog/member-key-vs-agent-key-vs-browser-sign-in) — 2026-10-07, by CoreSpeed. Four credentials, two kinds of principal. What each one reaches, where spend caps attach, and which credential an unattended agent should carry. (Markdown: https://corespeed.io/blog/member-key-vs-agent-key-vs-browser-sign-in.md)
- [Memory isolation with row-level security, not application code](https://corespeed.io/blog/memory-isolation-with-row-level-security) — 2026-10-07, by CoreSpeed. Caller identity comes from auth, each memory carries its scope, and Postgres row-level security enforces the boundary. Nothing in the request body can cross it. (Markdown: https://corespeed.io/blog/memory-isolation-with-row-level-security.md)
- [needs_reauth explained: what it means and what to do](https://corespeed.io/blog/needs-reauth-explained) — 2026-10-07, by CoreSpeed. needs_reauth means an account's upstream grant can no longer be used. Tools stay visible, calls fail, and the fix is reauthorizing in the dashboard. (Markdown: https://corespeed.io/blog/needs-reauth-explained.md)
- [OAuth connector vs pasting an API key: which to use for each app](https://corespeed.io/blog/oauth-connector-vs-pasting-an-api-key) — 2026-10-07, by CoreSpeed. Most CoreSpeed connectors use OAuth. Some take a pasted provider key, a few take only that. How the index tells them apart, and how rotation and reauth differ. (Markdown: https://corespeed.io/blog/oauth-connector-vs-pasting-an-api-key.md)
- [One MCP server vs a config file per agent](https://corespeed.io/blog/one-mcp-server-vs-a-config-file-per-agent) — 2026-10-07, by CoreSpeed. When a per-client list of local MCP servers is enough, where it stops scaling, and what changes once a team points every client at one remote server. (Markdown: https://corespeed.io/blog/one-mcp-server-vs-a-config-file-per-agent.md)
- [One org, many agents: Claude Code, Codex and a bot on one rulebook](https://corespeed.io/blog/one-org-many-agents) — 2026-10-07, by CoreSpeed. How a team keeps each member's own client, shares app connections, runs one approval policy and one ledger, and gives bots an identity of their own. (Markdown: https://corespeed.io/blog/one-org-many-agents.md)
- [payment_required and key_spend_limit_exceeded: what still works](https://corespeed.io/blog/payment-required-and-spend-limit-errors) — 2026-10-07, by CoreSpeed. payment_required is an organization wallet hold; key_spend_limit_exceeded is one key's monthly cap. Both arrive as HTTP 200 with isError true. (Markdown: https://corespeed.io/blog/payment-required-and-spend-limit-errors.md)
- [Pull requests for shared memory: proposals, archive and versions](https://corespeed.io/blog/pull-requests-for-shared-memory) — 2026-10-07, by CoreSpeed. A correction next to a stale shared memory does not stop the old one being retrieved. Archive it with a successor, or propose a fix for its author to merge. (Markdown: https://corespeed.io/blog/pull-requests-for-shared-memory.md)
- [Running an unattended job you operate with a CoreSpeed agent key](https://corespeed.io/blog/an-unattended-job-with-an-agent-key) — 2026-10-07, by CoreSpeed. Give a cron job or server process you run its own agent identity and sk-csa key: shared connections only, a monthly cap, approvals routed to its owner. (Markdown: https://corespeed.io/blog/an-unattended-job-with-an-agent-key.md)
- [Shared memory vs CLAUDE.md and AGENTS.md files](https://corespeed.io/blog/shared-memory-vs-instruction-files) — 2026-10-07, by CoreSpeed. Instruction files hold the rules an agent follows, per repository or per machine and per client. Org-owned memory holds searchable facts that follow you across both. Use both. (Markdown: https://corespeed.io/blog/shared-memory-vs-instruction-files.md)
- [Smart Approval vs client-side permission prompts](https://corespeed.io/blog/smart-approval-vs-client-side-permission-prompts) — 2026-10-07, by CoreSpeed. Claude Code's auto mode and Codex's Approve for me guard one client. Smart Approval gates the tools for every client. When each one is the right choice. (Markdown: https://corespeed.io/blog/smart-approval-vs-client-side-permission-prompts.md)
- [Social listening with read-only tools and no platform accounts](https://corespeed.io/blog/social-listening-with-read-only-tools) — 2026-10-07, by CoreSpeed. How an agent tracks mentions on X, Reddit, YouTube, TikTok and Xiaohongshu through read-only tools, with no platform login, and saves findings to shared memory. (Markdown: https://corespeed.io/blog/social-listening-with-read-only-tools.md)
- [The cs CLI in ten minutes: login, keys, tools, usage](https://corespeed.io/blog/the-cs-cli-in-ten-minutes) — 2026-10-07, by CoreSpeed. Install @corespeed/cs on Node 20 or Bun, sign in with cs login, mint and rotate API keys, print MCP config, search and call tools, check usage, and read exit codes in scripts. (Markdown: https://corespeed.io/blog/the-cs-cli-in-ten-minutes.md)
- [The hardest problem in multi-agent teams is scoping](https://corespeed.io/blog/the-hardest-problem-in-multi-agent-teams-is-scoping) — 2026-10-07, by CoreSpeed. Past one person, the hard question is scope: who may touch what, on whose behalf, with what budget, with what record. The organization has to answer it. (Markdown: https://corespeed.io/blog/the-hardest-problem-in-multi-agent-teams-is-scoping.md)
- [What are spend caps for AI agents, and how do they work?](https://corespeed.io/blog/what-are-spend-caps-for-ai-agents) — 2026-10-07, by CoreSpeed. Spend caps stop an agent loop from draining a budget. On CoreSpeed an org wallet threshold and per-key monthly caps refuse metered calls before a tool runs. (Markdown: https://corespeed.io/blog/what-are-spend-caps-for-ai-agents.md)
- [What does human-in-the-loop mean for AI agents in 2026?](https://corespeed.io/blog/what-is-human-in-the-loop-for-ai-agents) — 2026-10-07, by CoreSpeed. Human-in-the-loop means a person decides named actions before they run, under a policy written in advance. Reads never wait, and an approved call runs once. (Markdown: https://corespeed.io/blog/what-is-human-in-the-loop-for-ai-agents.md)
- [What is a plain-English approval policy?](https://corespeed.io/blog/what-is-a-plain-english-approval-policy) — 2026-10-07, by CoreSpeed. Sentences, read in full for every write an agent makes. Every save is a version, Test policy runs a hypothetical call, and a default you pick covers the gaps. (Markdown: https://corespeed.io/blog/what-is-a-plain-english-approval-policy.md)
- [What is a remote MCP server, and how do you connect one?](https://corespeed.io/blog/what-is-a-remote-mcp-server) — 2026-10-07, by CoreSpeed. A remote MCP server is reached over HTTPS instead of run as a local process. How sign-in works, and how to add one in Claude Code, Codex and Cursor. (Markdown: https://corespeed.io/blog/what-is-a-remote-mcp-server.md)
- [What is a shared connection, and when should you share one?](https://corespeed.io/blog/what-is-a-shared-connection) — 2026-10-07, by CoreSpeed. A shared connection is an app account every member and agent in the org can act through. Private is the default. Here is who can share, remove and reach one. (Markdown: https://corespeed.io/blog/what-is-a-shared-connection.md)
- [What is agent memory? Shared, private, and owned by the org](https://corespeed.io/blog/what-is-agent-memory) — 2026-10-07, by CoreSpeed. Agent memory is durable, searchable state that outlives one conversation, client or agent. How it differs from a context window, and who owns it. (Markdown: https://corespeed.io/blog/what-is-agent-memory.md)
- [What is an agent identity (and why not just an API key)?](https://corespeed.io/blog/what-is-an-agent-identity) — 2026-10-07, by CoreSpeed. An agent identity is a principal of its own, with an id, keys, an owner who answers for it, and reach limited to shared connections. A personal API key acts as you. (Markdown: https://corespeed.io/blog/what-is-an-agent-identity.md)
- [What is an audit trail for AI agents?](https://corespeed.io/blog/what-is-an-audit-trail-for-ai-agents) — 2026-10-07, by CoreSpeed. An audit trail records what each agent did, as whom, and how it ended. CoreSpeed keeps one per org: action, actor, outcome, code and cost, never the prompt. (Markdown: https://corespeed.io/blog/what-is-an-audit-trail-for-ai-agents.md)
- [What is an MCP gateway, and why do agents need one?](https://corespeed.io/blog/what-is-an-mcp-gateway) — 2026-10-07, by CoreSpeed. An MCP gateway is one remote MCP server that fronts many tools and app accounts behind one sign-in. What that changes for config, credentials, spend and audit. (Markdown: https://corespeed.io/blog/what-is-an-mcp-gateway.md)
- [What is credential custody for AI agents?](https://corespeed.io/blog/what-is-credential-custody-for-ai-agents) — 2026-10-07, by CoreSpeed. Credential custody means the agent asks for an action and the platform performs it with the right account. The agent never holds the token, so a leak exposes no app credential. (Markdown: https://corespeed.io/blog/what-is-credential-custody-for-ai-agents.md)
- [What is Harness-as-a-Service (HaaS)?](https://corespeed.io/blog/what-is-harness-as-a-service) — 2026-10-07, by CoreSpeed. Harness-as-a-Service is a hosted harness any agent signs into. Tools, auth, memory, budgets and an audit trail come as one service, and the agent you run stays yours. (Markdown: https://corespeed.io/blog/what-is-harness-as-a-service.md)
- [What is MCP, explained for teams running several agents](https://corespeed.io/blog/what-is-the-model-context-protocol-for-teams) — 2026-10-07, by CoreSpeed. MCP is the JSON-RPC protocol an agent uses to list and call tools on a server. For teams, one remote server means shared connections, one policy and one ledger. (Markdown: https://corespeed.io/blog/what-is-the-model-context-protocol-for-teams.md)
- [What is SKILL.md, and why is setup one line?](https://corespeed.io/blog/what-is-skill-md-and-the-one-line-setup) — 2026-10-07, by CoreSpeed. SKILL.md is the root file of an agent skill: a name, a description and instructions. CoreSpeed publishes its setup as one, so a single chat line installs it. (Markdown: https://corespeed.io/blog/what-is-skill-md-and-the-one-line-setup.md)
- [When to bring your own OAuth app (Google, Microsoft, Meta) and how](https://corespeed.io/blog/bring-your-own-oauth-app) — 2026-10-07, by CoreSpeed. Google, Microsoft and Meta gate shared OAuth apps behind review. An org admin registers your own app: paste client_id and client_secret, match the scopes. (Markdown: https://corespeed.io/blog/bring-your-own-oauth-app.md)
- [Why is a tool missing from tools/list?](https://corespeed.io/blog/why-is-a-tool-missing-from-tools-list) — 2026-10-07, by CoreSpeed. A tool is missing from tools/list because this caller cannot call it now: app not connected, capability off, agent key reach, or a stale session. (Markdown: https://corespeed.io/blog/why-is-a-tool-missing-from-tools-list.md)
- [Why reads are never gated: designing an approval gate people keep on](https://corespeed.io/blog/reads-are-never-gated) — 2026-10-07, by CoreSpeed. A gate that asks about everything gets switched off. Smart Approval judges writes only, lets reads through, and never lets a failed evaluation run a write. (Markdown: https://corespeed.io/blog/reads-are-never-gated.md)
- [Why tools/list is yours, and not a catalog](https://corespeed.io/blog/what-is-tools-list-and-why-it-is-yours) — 2026-10-07, by CoreSpeed. tools/list on CoreSpeed returns exactly what this sign-in can call. Absence is a decision, presence is neither health nor permission, so never hard-code it. (Markdown: https://corespeed.io/blog/what-is-tools-list-and-why-it-is-yours.md)
- [Why your agent should never hold an OAuth token](https://corespeed.io/blog/why-your-agent-should-never-hold-an-oauth-token) — 2026-10-07, by CoreSpeed. An agent that holds a token can leak it through a prompt, a log or a transcript. Credential custody lets a platform act with the right account instead. (Markdown: https://corespeed.io/blog/why-your-agent-should-never-hold-an-oauth-token.md)
- [Introducing Smart Approval](https://corespeed.io/blog/introducing-smart-approval) — 2026-09-25, by Hao Su. Decide which of your agents' actions need your sign-off. Everything else keeps running; the ones you named wait for you in Slack, by email or on the dashboard. (Markdown: https://corespeed.io/blog/introducing-smart-approval.md)
- [How to turn CoreSpeed into a content factory](https://corespeed.io/blog/how-to-turn-corespeed-into-a-content-factory) — 2026-09-17, by Spencer Zhao. How we made PRODUCTION INCIDENT, from Blender rehearsal and generated performances to screen compositing, sound, and the final 4K edit. (Markdown: https://corespeed.io/blog/how-to-turn-corespeed-into-a-content-factory.md)
- [A letter from our founders](https://corespeed.io/blog/a-letter-from-our-founders) — 2026-09-14, by Milton Yan, Hao Su. Why we're building an operating system for agents, and what becomes possible when they have one. (Markdown: https://corespeed.io/blog/a-letter-from-our-founders.md)
- [Introducing CoreSpeed](https://corespeed.io/blog/introducing-corespeed) — 2026-09-14, by Hao Su, Milton Yan, Spencer Zhao, Benji Huang. An operating system for agents. (Markdown: https://corespeed.io/blog/introducing-corespeed.md)
- [How we took LongMemEval from 80.0 to 94.2 without touching retrieval](https://corespeed.io/blog/how-we-took-longmemeval-from-80-to-94-without-touching-retrieval) — 2026-08-27, by Spencer Zhao. LongMemEval-S climbed from 80.0 to 94.2 while retrieval recall at 10 stayed at 0.99. The gains came from evidence budget, reader generation, and instructions. (Markdown: https://corespeed.io/blog/how-we-took-longmemeval-from-80-to-94-without-touching-retrieval.md)
- [We retrieved the memory, then dropped the answer](https://corespeed.io/blog/we-retrieved-the-memory-then-dropped-the-answer) — 2026-08-22, by Spencer Zhao. A 200-question conflict run scored 0.800 at the parent-Memory boundary and 0.635 at the evidence boundary. The gap was in what reached the reader. (Markdown: https://corespeed.io/blog/we-retrieved-the-memory-then-dropped-the-answer.md)

RSS: https://corespeed.io/blog/rss.xml