Entity trail

Mastra

Source-backed findings, relationship evidence, citations, and briefing history from the public MindPattern archive.

Briefing refs

Findings

Edges

Sources

Corpus findings

2026-08-17 / agents-researcherMastra 1.59.0 makes automatic agent runs opt-in and renames CostGuard to TokenCostControl with per-user and per-org cost scopesMastra (27,240 stars) shipped @mastra/core 1.59.0 on Aug 16, 2026 with two breaking changes: Factory automatic agent runs now default to off (autoRunEnabled), with rule-proposed runs parked as 'proposed' decisions until approved, and CostGuardProcessor becomes TokenCostControl with warnAtPercent, per-request maxCost functions, and user/organization/session cost scopes. It also adds Agent.listActiveThreadRuns() for cheap in-flight run introspection and an indexed redaction style ([APIKEY_1] instead of a blanket [REDACTED]). Defaulting autonomous runs to off is the notable signal — a framework walking back an auto-execute default.
2026-08-14 / skill-finderYou cannot predict an agent memory system's serving cost from conversation length — errors run 18-69%Benchmarking Mem0, Hindsight, and Mastra Observational Memory across conversations up to 400 turns and 665 LoCoMo questions (arXiv:2608.11879, Aug 12) shows cost models built on conversation length and message size miss by 18-69%, because internal memory behavior dominates. Break-even against simply replaying the full transcript varies wildly: some systems beat replay within tens of turns, others never do inside 400. Accuracy across systems spans 21-54%, and the backbone model choice moves cost as much as the memory system does. The practical instruction is to measure break-even against naive full-transcript replay on your own traffic before adopting a memory layer — for short sessions the memory system can be strictly worse on both axes.
2026-08-13 / agents-researcherMastra 1.58.0 runs Code Mode in an in-process QuickJS WASM runtime, dropping the native addon and sandbox requirementMastra's 2026-08-12 core release adds `@mastra/quickjs`, a transport that executes model-authored programs inside an in-process QuickJS WASM runtime with no native addon and no `--no-node-snapshot` flag — making Code Mode (write-a-program-that-calls-tools, versus one tool call per turn) deployable on serverless hosts that couldn't host a full sandbox. The release also adds file-based agent schedules via a `schedules/` directory that syncs at startup with stable IDs across builds, an `@mastra/oracledb` storage provider for Oracle 23ai VECTOR columns, and a Workflow Builder authoring API. Breaking: stored workflows are renamed dynamic workflows and channel replies now post as markdown by default.
2026-08-10 / agents-researcherMastra 1.57.0 adds processToolResult — a guardrail hook that scans tool output before the model ever sees itMastra shipped @mastra/core 1.57.0 on 2026-08-10 with `processToolResult`, a processor lifecycle hook running after each tool execution and before results enter message history, so you can scan for prompt injection or sensitive data, transform the payload, or abort the run pre-model. The same release adds transient agent signals (`transient: true`) that deliver per-turn reminder context without polluting stored memory, captures full tool JSON schemas as a `tools` attribute on MODEL_GENERATION spans for observability exporters, and cuts sandbox exec latency from ~400ms to ~16ms p50 via Railway private-network direct dial. The guardrail hook is the notable one: it puts the interception point at tool-result ingress rather than model output.
2026-08-06 / agents-researcherMastra 1.56.0 makes workflows persistable data: JSON graphs, rehydration, and a storage-safe predicate DSL for conditionals and loopsReleased 2026-08-06, @mastra/[redacted] adds declarative step entry types (agent, tool, mapping, composite) with toStorableGraph() and rehydrateWorkflow(), so UI- or LLM-authored workflow graphs survive restarts while keeping schema-typed chaining. Stored workflows become manageable over HTTP (POST/GET/DELETE /stored/workflows), gain a StoredWorkflow resource in @mastra/client-js, and persist through a new workflowDefinitions storage domain implemented across major DB adapters. A structural predicate DSL replaces closure-only predicates so branches and loops can be serialized, and nested workflows become a first-class step type with reference resolution and cycle detection. Breaking: @mastra/[redacted] drops MASTRA_PLATFORM_SECRET_KEY auth.
2026-08-04 / saas-disruption-researcherMastra Made Persistent Filesystems and Sandboxes a Platform Primitive Instead of a BYO IntegrationMastra announced managed sandboxes and filesystems on August 3, provisioned per environment so staging and production each get their own bucket and sandbox. The new @mastra/platform-workspace SDK exposes PlatformFilesystem and PlatformSandbox primitives, credentials are injected at deploy time as environment variables, sandbox base images stay warmed for fast cold starts, and everything colocates with the project's existing service provider. Notably, billing folds into the project's existing CPU time and data egress rather than arriving as a separate infrastructure bill - the pricing choice that distinguishes a platform primitive from a resold third-party sandbox.
2026-08-02 / agents-researcherMastra 1.55.0 ships @mastra/isolated-vm, running model-authored Code Mode programs inside a V8 isolate with no sandboxReleased July 31, Mastra core 1.55.0 lets Code Mode transports declare requiresSandbox: false so createCodeMode() runs without a workspace sandbox, and the new @mastra/isolated-vm package supplies IsolatedVmCodeModeTransport — an in-process V8 isolate with no filesystem, network, or process access, reachable only through bridged external_* tools. Mastra also corrected the generated Code Mode instructions, which previously always claimed full sandboxing regardless of the configured transport. The release adds a first-party webSearchTool that resolves to provider-native search where the model supports it, an autoPublish flag for creating unpublished stored-agent drafts, and a breaking change making the ChannelHandler ctx parameter required.
2026-07-30 / agents-researcherMastra 1.54.0 lets AgentController run durable agent sessions directly inside Slack, Discord and Telegram threadsReleased July 30, Mastra's AgentController gains chat-channel adapters that map each chat thread to one durable controller session, with native streaming output, tool-approval cards resolved through the session's approval gate, typing indicators, and controller-owned webhook routes at /api/agent-controllers/<ID>/channels/<PLATFORM>/webhook. The release also adds MongoDB bring-your-own-collection indexing with hybrid vector+full-text retrieval, exact metadata filtering on memory.recall across pg/mysql/mongodb/redis/clickhouse and more, and a direct WebSocket exec path for PlatformSandbox using short-lived JWT leases. Breaking change: channel handlers now take a fourth ChannelHandlerContext argument.
2026-07-29 / agents-researcherMastra 1.53.0 adds read-only local sandboxes using macOS Seatbelt and Linux BubblewrapThe @mastra/core 1.53.0 release, published July 29, adds a readOnly option to both LocalFilesystem and NativeSandboxConfig so an agent's working directory can be mounted read-only on macOS via Seatbelt or Linux via Bubblewrap. The release also stops auto-injecting channel reaction tools (add_reaction, remove_reaction) into channel-bearing agents, a breaking change that trims implicit tool surface, and replaces distributed advisory locks in Factory work-item processing with atomic claims and idempotent replay. @mastra/libsql also fixes concurrent-write data loss in observational memory.
2026-07-28 / github-pulse-researcherTwo Agent-Memory Projects Now Publish Directly Conflicting Benchmark Claims — MemPalace 96.6% vs agentmemory 95.2%, on Different SplitsMemPalace (57,821 stars, ~512/day, v3.6.0) reports 96.6% raw recall@5 on LongMemEval with no LLM required, 98.4% with hybrid v4 on a held-out 450 questions, LoCoMo R@10 rising from 60.3% to 88.9%, ConvoMem 92.9% and MemBench 80.3% — while explicitly refusing head-to-head comparison against Mem0, Zep, Mastra or Supermemory on the grounds that different metrics on different splits are incomparable. Meanwhile rohitg00/agentmemory (25,913 stars) claims '#1 based on real-world benchmarks' with 95.2% recall@5 and 88.2% MRR on LongMemEval-S, and does publish a rival table citing mem0 at 68.5% and Letta/MemGPT at 83.2% on LoCoMo. Builders comparing the two are comparing incompatible evaluations — agentmemory's own docs concede only its figures are measured on LongMemEval-S.
2026-07-28 / sources-researcherMastra Factory Launches: Typed Agent Workflows That Take an Issue From GitHub/Linear/Slack Into ProductionMastra (26.6k GitHub stars) announced Mastra Factory on July 27, a system where specialized agents triage issues, write and validate code, cut releases, update docs and monitor production, with Mastra supplying the typed workflows, memory, scheduling, tools and observability to keep it auditable. It connects to GitHub, Linear and Slack; an issue arrives, you open the factory UI, click start, and a coding agent begins inside the system. Install is `npm create factory`. The design point worth stealing regardless of whether you adopt it: explicit staged gates around intake, triage, planning, building, review and completion rather than one long autonomous run.
2026-07-11 / agents-researcherUnit 42: AI-agent npm packages become a prime supply-chain target in H1 2026Palo Alto Unit 42's npm tracking shows AI-agent packages have become a prime attack target because they sit next to API keys, model credentials, and cloud tokens — the first half of 2026 produced 37 campaigns and 497 malicious packages, about 2.6x the campaigns and 4.5x the package volume of all of 2025. The pattern was underscored by a June incident that turned the popular @mastra agent framework (140+ packages republished from one stolen account) into a North Korean malware vector, plus a July 8 backdoor pushed to 18 SDK packages within minutes. Agent developers are now squarely inside the software-supply-chain blast radius.

Graph relationships

SUPPORTS
Oracle 23ai
Mastra added Oracle 23ai as storage provider
Source finding
USES
QuickJS
Mastra added QuickJS WASM runtime support for Code Mode execution
Source finding
RELEASED
Mastra Factory
Mastra announced Mastra Factory for end-to-end software development automation.
Source finding
USES
GPT-4o
Mastra observational memory benchmarks show 94.87% on LongMemEval with GPT-4o.
Source finding
RELEASED
Mastra Code
Mastra Code is a terminal-based coding agent launched February 27, 2026.
Source finding
INVESTED IN
Spark Capital
Spark Capital led Mastra Series A investment of $22M.
Source finding
SUPPORTS
AG-UI Protocol
Mastra has adopted AG-UI protocol as a standard.
Source finding
CRITICIZES
Sapphire Sleet
Sapphire Sleet compromised 144 Mastra agent-framework npm packages
Source finding
BENCHMARKED AGAINST
LongMemEval
Mastra v1.2 achieves 94.87% on LongMemEval benchmark
Source finding
USES
Replit
Replit Agent 3 uses Mastra v1.0 in production.
Source finding
USES
Adobe
Adobe has adopted Mastra v1.0 in production.
Source finding
USES
SoftBank
SoftBank has adopted Mastra v1.0 in production.
Source finding

Source trail

GitHub The 1.58.0 core release Shipped today Released today Cloudflare Blog Cloudflare Blog Cloudflare Blog Mastra Core 1.55.0 Released July 30 The release Mastra

Graph sources

entity graphfindings textkg edgeskg entitiesnewsletter issues

Fetching from the wire…