This release adds 13 notable features for engineering teams evaluating rollout.
✓ No known CVEs patched in this version
Topics
Summary
AI summaryPerformance Improvements, code updates, and Bug Fixes were made.
Changes in this release
| Type | Severity | Summary | CVE |
|---|---|---|---|
| Feature | Medium |
Adds Meta model provider. Adds Meta model provider. Source: llm_adapter@2026-07-14 Confidence: high |
— |
| Feature | Medium |
Adds /tools slash command. Adds /tools slash command. Source: llm_adapter@2026-07-14 Confidence: high |
— |
| Feature | Medium |
Adds configurable chat cursor style. Adds configurable chat cursor style. Source: llm_adapter@2026-07-14 Confidence: high |
— |
| Feature | Medium |
Adds in‑app Debug Console. Adds in‑app Debug Console. Source: llm_adapter@2026-07-14 Confidence: high |
— |
| Feature | Medium |
Adds `-s` alias for `--skill`. Adds `-s` alias for `--skill`. Source: llm_adapter@2026-07-14 Confidence: high |
— |
| Feature | Medium |
Adds support for Fireworks `/routers` model IDs. Adds support for Fireworks `/routers` model IDs. Source: llm_adapter@2026-07-14 Confidence: high |
— |
| Feature | Medium |
Adds `/model` Ctrl+N toggle for names vs raw specs. Adds `/model` Ctrl+N toggle for names vs raw specs. Source: llm_adapter@2026-07-14 Confidence: high |
— |
| Feature | Medium |
Adds abort option in `/threads` cwd‑switch prompt. Adds abort option in `/threads` cwd‑switch prompt. Source: llm_adapter@2026-07-14 Confidence: high |
— |
| Feature | Medium |
Adds resume threads in‑TUI with `/threads -r [ID]`. Adds resume threads in‑TUI with `/threads -r [ID]`. Source: llm_adapter@2026-07-14 Confidence: high |
— |
| Feature | Low |
Sets `prompt_cache_key` for OpenAI models. Sets `prompt_cache_key` for OpenAI models. Source: granite4.1:30b@2026-07-14-audit Confidence: low |
— |
| Feature | Low |
Expands environment variables in MCP configuration. Expands environment variables in MCP configuration. Source: granite4.1:30b@2026-07-14-audit Confidence: low |
— |
| Feature | Low |
Hides diff widget for credential files. Hides diff widget for credential files. Source: granite4.1:30b@2026-07-14-audit Confidence: low |
— |
| Feature | Low |
Shows “(debug enabled)” on splash when `DEEPAGENTS_CODE_DEBUG` is set. Shows “(debug enabled)” on splash when `DEEPAGENTS_CODE_DEBUG` is set. Source: granite4.1:30b@2026-07-14-audit Confidence: low |
— |
| Performance | Medium |
Improves performance by loading MCP servers concurrently during graph build. Improves performance by loading MCP servers concurrently during graph build. Source: llm_adapter@2026-07-14 Confidence: high |
— |
| Bugfix | Medium |
Avoids repeated startup auto‑update stalls. Avoids repeated startup auto‑update stalls. Source: granite4.1:30b@2026-07-14-audit Confidence: low |
— |
| Bugfix | Low |
Infers Fireworks provider from qualified model IDs. Infers Fireworks provider from qualified model IDs. Source: granite4.1:30b@2026-07-14-audit Confidence: low |
— |
| Bugfix | Low |
Captures input typed before TUI startup. Captures input typed before TUI startup. Source: granite4.1:30b@2026-07-14-audit Confidence: low |
— |
| Bugfix | Low |
Detaches owned `langgraph dev` server from terminal. Detaches owned `langgraph dev` server from terminal. Source: granite4.1:30b@2026-07-14-audit Confidence: low |
— |
| Bugfix | Low |
Infers additional model providers. Infers additional model providers. Source: granite4.1:30b@2026-07-14-audit Confidence: low |
— |
| Bugfix | Low |
Preserves `Ctrl+D` deletion in non‑empty input. Preserves `Ctrl+D` deletion in non‑empty input. Source: granite4.1:30b@2026-07-14-audit Confidence: low |
— |
| Bugfix | Low |
Allows quitting with `Ctrl+D` at end of prompt. Allows quitting with `Ctrl+D` at end of prompt. Source: granite4.1:30b@2026-07-14-audit Confidence: low |
— |
| Bugfix | Low |
Reaps langgraph dev server when startup is cancelled. Reaps langgraph dev server when startup is cancelled. Source: granite4.1:30b@2026-07-14-audit Confidence: low |
— |
| Bugfix | Low |
Rejects `--auto-approve` in headless mode. Rejects `--auto-approve` in headless mode. Source: granite4.1:30b@2026-07-14-audit Confidence: low |
— |
| Bugfix | Low |
Routes explicit `--stdin` + `--skill` to headless path. Routes explicit `--stdin` + `--skill` to headless path. Source: granite4.1:30b@2026-07-14-audit Confidence: low |
— |
| Bugfix | Low |
Skips Esc prompt‑restore once output generation begins. Skips Esc prompt‑restore once output generation begins. Source: granite4.1:30b@2026-07-14-audit Confidence: low |
— |
| Bugfix | Low |
Removes misleading agent names from help output. Removes misleading agent names from help output. Source: granite4.1:30b@2026-07-14-audit Confidence: low |
— |
| Bugfix | Low |
Supports plain `exit` quit command. Supports plain `exit` quit command. Source: granite4.1:30b@2026-07-14-audit Confidence: low |
— |
| Bugfix | Low |
Syncs `ask_user` active‑question highlight with focus. Syncs `ask_user` active‑question highlight with focus. Source: granite4.1:30b@2026-07-14-audit Confidence: low |
— |
| Bugfix | Low |
Wraps MCP viewer navigation. Wraps MCP viewer navigation. Source: granite4.1:30b@2026-07-14-audit Confidence: low |
— |
Full changelog
Features
- Add Meta model provider (#4650) (70829c5)
- Set
prompt_cache_keyfor OpenAI models (#4632) (8cf57ac) - Support Fireworks
/routersmodel ids (#4591) (1c08d27) /modelCtrl+N toggle for names vs raw specs (#4592) (518c322)/toolsslash command (#4649) (b1600a8)- Add
-salias for--skill(#4620) (c9b7ac2) - Configurable chat cursor style (#4687) (a22484b)
- Expand environment variables in MCP config (#4681) (4f5d7be)
- Hide diff widget for credential files (#4593) (4c49a24)
- In-app Debug Console (#4564) (4f94a30)
- Offer abort in
/threadscwd-switch prompt (#4583) (aaeac99) - Resume threads in-TUI with
/threads -r [ID](#4609) (d442673) - Show
(debug enabled)on splash whenDEEPAGENTS_CODE_DEBUGis set (#4584) (f10b877)
Bug Fixes
- Avoid repeated startup auto-update stalls (#4648) (12a9c9d)
- Infer Fireworks provider from qualified model IDs (#4594) (4d2aa8a)
- Capture input typed before TUI startup (#4684) (ef9a4a8)
- Detach owned
langgraph devserver from terminal (#4642) (d1f3afe) - Infer additional model providers (#4675) (4ceed24)
- Preserve
Ctrl+Ddeletion in non-empty input (#4626) (306bd89) - Quit with
Ctrl+Dat end of prompt (#4678) (2f8c111) - Reap langgraph dev server when startup is cancelled (#4629) (904ff05)
- Reject
--auto-approvein headless mode (#4617) (997be16) - Route explicit
--stdin+--skillto headless path (#4611) (724e24a) - Skip Esc prompt-restore once output generation begins (#4582) (14f384f)
- Remove misleading agent names from help (#4671) (ac15732)
- Support plain
exitquit command (#4543) (e6f10a1) - Sync
ask_useractive-question highlight with focus (#4599) (e4c29b5) - Wrap MCP viewer navigation (#4677) (cffc732)
Performance Improvements
Internal maintainers: @johannes117, @wbbradley, @agola11, @ccurme, @mdrxy
Released by: @mdrxy
Git log since deepagents-code==0.1.36This commit history includes changes to this package. Commits are listed newest first.
def6369release(deepagents-code): 0.1.37 (#4619)4f5d7befeat(code): expand environment variables in MCP config (#4681)a22484bfeat(code): add configurable chat cursor style (#4687)ef9a4a8fix(code): capture input typed before TUI startup (#4684)e4c29b5fix(code): syncask_useractive-question highlight with focus (#4599)724e24afix(code): route explicit--stdin+--skillto headless path (#4611)cffc732fix(code): wrap MCP viewer navigation (#4677)b2d62cehotfix(code): restoreShift+Tabin debug console (#4680)4ceed24fix(code): infer additional model providers (#4675)2f8c111fix(code): quit withCtrl+Dat end of prompt (#4678)b1600a8feat(code):/toolsslash command (#4649)4f94a30feat(code): in-app Debug Console (#4564)12a9c9dfix(code): avoid repeated startup auto-update stalls (#4648)306bd89fix(code): preserveCtrl+Ddeletion in non-empty input (#4626)e6f10a1fix(code): support plainexitquit command (#4543)8cf57acfeat(code): setprompt_cache_keyfor OpenAI models (#4632)70829c5feat(code): add Meta model provider (#4650)aaeac99feat(code): offer abort in/threadscwd-switch prompt (#4583)4d2aa8afix(code): infer Fireworks provider from qualified model IDs (#4594)4c49a24feat(code): hide diff widget for credential files (#4593)d1f3afecfix(code): detach ownedlanggraph devserver from terminal (#4642)f10b877feat(code): show(debug enabled)on splash whenDEEPAGENTS_CODE_DEBUGis set (#4584)d442673feat(code): resume threads in-TUI with/threads -r [ID](#4609)ac15732fix(code): remove misleading agent names from help (#4671)14f384ffix(code): skip Esc prompt-restore once output generation begins (#4582)c5345ccperf(code): load MCP servers concurrently during graph build (#4659)1c08d27feat(code): support Fireworks/routersmodel ids (#4591)518c322feat(code):/modelCtrl+N toggle for names vs raw specs (#4592)c9b7ac2feat(code): add-salias for--skill(#4620)e9c6130chore(code,evals,talon): bumplanggraph-runtime-inmemto 0.31.0 (#4647)c87e8fedocs(code): warn thatsystem_promptoverride bypasses generated prompt (#4644)904ff05fix(code): reap langgraph dev server when startup is cancelled (#4629)ddce7c3chore(code): reframe dcode pyproject description as coding agent (#4610)997be16fix(code): reject--auto-approvein headless mode (#4617)
Weekly OSS security release digest.
The CVE patches and breaking changes that affected production tools this week. One email, every Sunday.
No spam, unsubscribe anytime.
Share this release
About deepagents
Agent harness built with LangChain and LangGraph. Equipped with a planning tool, a filesystem backend, and the ability to spawn subagents - well-equipped to handle complex agentic tasks.
Related context
Related tools
Earlier breaking changes
- vlangchain-quickjs==0.2.0 Removes `skills_backend` capability.
- vdeepagents-cli==0.2.0 `deepagents deploy` now targets Managed Deep Agents API; removes `deepagents dev`, `deepagents.toml`, and `--config`/`--dry-run` flags.
Beta — feedback welcome: [email protected]